Mechanistic insights into c-di-GMP–dependent control of the biofilm regulator FleQ from Pseudomonas aeruginosa

Significance

Pseudomonas aeruginosa, an opportunistic pathogen that can cause fatal chronic infections, relies on the intracellular second-messenger c-di-GMP to form robust multicellular biofilms during host tissue colonization. c-di-GMP is sensed directly by the transcription regulator FleQ, which inversely regulates flagellar motility and exopolysaccharide secretion to secure a planktonic to sessile life-form transition. FleQ belongs to the diverse family of AAA+ ATPase enhancer-binding proteins, but how its noncanonical function on transcriptional regulation is controlled by c-di-GMP remains enigmatic. Here, we report structural and functional data that identify an unusual mode of c-di-GMP recognition accompanied by a major quaternary structure reorganization. Our analyses offer a consensus to previous studies and unique insights into the mechanism of action of FleQ and FleQ-like proteins.

Abstract

Bacterial biofilm formation during chronic infections confers increased fitness, antibiotic tolerance, and cytotoxicity. In many pathogens, the transition from a planktonic lifestyle to collaborative, sessile biofilms represents a regulated process orchestrated by the intracellular second-messenger c-di-GMP. A main effector for c-di-GMP signaling in the opportunistic pathogen Pseudomonas aeruginosa is the transcription regulator FleQ. FleQ is a bacterial enhancer-binding protein (bEBP) with a central AAA+ ATPase σ⁵⁴-interaction domain, flanked by a C-terminal helix-turn-helix DNA-binding motif and a divergent N-terminal receiver domain. Together with a second ATPase, FleN, FleQ regulates the expression of flagellar and exopolysaccharide biosynthesis genes in response to cellular c-di-GMP. Here we report structural and functional data that reveal an unexpected mode of c-di-GMP recognition that is associated with major conformational rearrangements in FleQ. Crystal structures of FleQ’s AAA+ ATPase domain in its apo-state or bound to ADP or ATP-γ-S show conformations reminiscent of the activated ring-shaped assemblies of other bEBPs. As revealed by the structure of c-di-GMP–complexed FleQ, the second messenger interacts with the AAA+ ATPase domain at a site distinct from the ATP binding pocket. c-di-GMP interaction leads to active site obstruction, hexameric ring destabilization, and discrete quaternary structure transitions. Solution and cell-based studies confirm coupling of the ATPase active site and c-di-GMP binding, as well as the functional significance of crystallographic interprotomer interfaces. Taken together, our data offer unprecedented insight into conserved regulatory mechanisms of gene expression under direct c-di-GMP control via FleQ and FleQ-like bEBPs.

Bacterial adaptations to diverse environments, including human hosts, involve collaborative group behaviors, such as quorum sensing, swarming, and biofilm formation (1–5). In general, quorum-sensing during host tissue colonization is associated with virulence gene expression and acute-phase infections, whereas biofilm formation facilitates the development of chronic infections, evasion of host immune response, and increased tolerance to treatments (6). It is now well appreciated that these social behaviors result from highly regulated signal transduction processes, which in many bacteria are choreographed by the nucleotide-based second messenger c-di-GMP (7–9). Synthesized by GGDEF domain-containing diguanylate cyclases and hydrolyzed by EAL or HD-GYP domain-containing phosphodiesterases, c-di-GMP is sensed by a variety of protein- and RNA-based effectors to exert control at transcriptional, translational, and posttranslational levels (10, 11).

In Pseudomonas aeruginosa, an opportunistic pathogen that causes severe chronic infections in cystic fibrosis patients, burn victims, and other immunocompromised individuals, the transcription factor FleQ acts as a master regulator of flagellar motility and exopolysaccharide (EPS) secretion to mediate the transition between planktonic and biofilm lifestyles (12–14). FleQ contains an N-terminal REC receiver domain (Pa_FleQ^REC), a central AAA+ ATPase σ⁵⁴-interaction domain (Pa_FleQ^AAA), and a C-terminal helix-turn-helix DNA binding motif (Pa_FleQ^HTH) (Fig. 1A). FleQ shares significant homology and similarities in gene-expression regulation with members of the NtrC subfamily of bacterial enhancer binding proteins (bEBPs) (15).

Fig. 1.

Crystal structures of P. aeruginosa FleQ AAA+ domain reveal ring-shaped, hexameric assemblies. (A) Domain organization of FleQ. AAA+ functional motifs and bEBPs specific features discussed in the text are labeled. Full-length P. aeruginosa FleQ (Pa_FleQ^Full) and domain construct boundaries (Pa_FleQ^AAA and Pa_FleQ^AAA-HTH) used in this work are indicated. (B) Structure of an ATP-γ-S-bound Pa_FleQ^AAA protomer. Coloring of structural elements and AAA+ subdomains SD1 and SD2 is consistent with the diagram shown in A. (C) Superimposition of apo (gray), ADP-bound (pink), and ATP-γ-S-bound (orange) crystal structures. Red and brown dashed lines highlight interactions between FleQ residues and ADP or ATP-γ-S, respectively. (D) Crystallographic apo-Pa_FleQ^AAA hexamer assembly as viewed down the 6₅ fold axis. Symmetry generated protomers are colored in black. Walker A motif and σ⁵⁴-interacting loops L1 and L2 are shown in blue, magenta and cyan, respectively. (Right) An extended view of the crystal lattice in an orthogonal view. Intercalated black and white “open” hexamers appear as helical Pa_FleQ^AAA filaments running along the screw axis through the entire crystals.

Typically, bEBP with REC domains are regulated by phosphorylation via cognate histidine kinases, whereby transfer of a phosphoryl group from the kinase to the REC domain alters molecular interactions within or between bEBPs (16, 17). As a result, phosphorylated bEBPs, reorganized into ring-shape hexamers, bind ∼150 bp upstream of the transcription start site, recruit the σ⁵⁴-RNAP–promoter complex, and initiate transcription in an ATPase-dependent manner (18–20). Despite similar domain architecture, however, FleQ appears to use a drastically different mechanism of regulation (13, 14, 21). FleQ lacks both a cognate sensor kinase and conserved REC domain residues that are crucial for phosphotransfer and phosphorylation-dependent conformational changes. Instead of relying on covalent modifications, its function depends on direct protein–protein interactions with FleN, a distinct ATPase that acts largely as a FleQ antagonist (22). Furthermore, at promoters of the flagellar biosynthesis gene cluster, FleQ not only binds to enhancer sequences upstream of the transcription start site—like classic bEBPs—but also at sites downstream from the transcription start sites (13, 23). In addition, and most relevant for this study, c-di-GMP directly binds to and regulates FleQ, altering target promoter activity (13, 14, 21).

C-di-GMP exerts distinct effects on FleQ’s regulatory roles at the flagellar and EPS biosynthesis operons, underscoring the system’s multifaceted role in biofilm formation (21). The dinucleotide dampens FleQ’s ATPase activity and causes a modest but significant down-regulation of flagellar gene expression (13, 14, 21). In contrast, FleQ acts as a repressor of Pel EPS expression in the absence of c-di-GMP. An increase of intracellular c-di-GMP levels through the action of diguanylate cyclases not only relieves FleQ-mediated gene repression, but it converts the protein into a potent transcription activator (13). Strikingly, gene expression activation under these conditions does not appear to require FleQ’s intrinsic ATPase activity, is σ⁵⁴-independent, and likely proceeds through σ⁷⁰ (21, 24). On both flagellar and pel promoters FleN further modulates FleQ-mediated gene regulation (14, 21). However, FleN’s effect on the regulation of flagella expression and assembly is much more pronounced, with deletion of P. aeruginosa fleN leading to multiflagellated cells and only a small down-regulation in pel expression (14, 25).

How FleQ is able to exert such different effects and to sense different regulatory inputs at its target promoters has so far remained enigmatic. With regard to c-di-GMP signal recognition, at least two distinct modes of second-messenger binding in the context of FleQ’s tertiary structure have been proposed (21, 26). It is unclear, however, how c-di-GMP recognition would affect the various aspects of the protein’s higher-order regulatory mechanics. Here we present crystallographic data and solution-based structure-function analyses that reveal an unexpected mode of c-di-GMP regulation. In particular, we show that binding of the second messenger occurs at a composite site within the FleQ AAA+ domain and is accompanied by a major quaternary structure reorganization and allosteric ATPase inhibition. Our results reconcile previous modeling and functional studies (13, 14, 21, 26), and provide an integrated model for FleQ-dependent transcription regulation of genes involved in EPS production and secretion.

Results

Canonical, Ring-Shaped FleQ Hexamers Revealed by Crystal Structures of Its AAA+ Domain.

The AAA+ ATPase domain of bEBPs couples ATP binding and hydrolysis to transcription complex remodeling (18). All bEBPs’ functional features, including the specific σ⁵⁴-interaction loops—L1, which contains the signature GAFTGA motif (27), and L2—are conserved in FleQ (Fig. 1A). Purified, nucleotide-free Pa_FleQ^AAA (residues 137–394) crystallized in space group P6₅ and its structure was determined to 1.8 Å resolution by single-wavelength anomalous dispersion (SAD) phasing on iodide-derivatized crystals. Using the apo-structure as the search model, structures of Pa_FleQ^AAA cocrystallized with ADP or ATP-γ-S were solved by molecular replacement and refined to a resolution of 2.4 Å and 2.5 Å, respectively (Table S1). In these structures, Pa_FleQ^AAA adopts a canonical bilobal fold comprising a larger N-terminal α/β subdomain typical for P-loop NTPases (SD1) and an α-helical subdomain specific for AAA+ ATPases (SD2) (Fig. 1B) (20, 28). Apart from subtle changes caused by nucleotide coordination, the apo- and nucleotide-bound structures are virtually identical, with an all-atom rmsd lower than 0.6 Å in pairwise superpositions of protomers. Both ADP and ATP-γ-S interact with residues S¹⁷⁶ and K¹⁸⁰ from Walker A (G¹⁷⁴xxGxGKE), as well as with a conserved R³⁶³ from the so-called sensor II motif in subdomain SD2 (Fig. 1C and Fig. S1D) (28). Ribose sugar moieties are additionally coordinated by the SD2 residue R³³⁴. In the ATP-γ-S–bound form, the γ-phosphate of the nucleotide is sensed by D²⁴⁵ and, indirectly, by E²⁴⁶ from the Walker B motif (T²⁴¹LFLDE) (Fig. 1C and Fig. S1A).

Table S1.

Data collection and refinement statistics

Data collection and refinement	Pa_FleQAAA NaI derivative	Pa_FleQAAA	Pa_FleQAAA+ ADP	Pa_FleQAAA+ ATP-γ-S	Pa_FleQAAA-HTH + c-di-GMP
Data collection
Wavelength (Å)	1.4592	1.4592	1.4592	1.5418	0.9792
Space group	P65	P65	P65	P65	P6322
Unit cell (Å)	104.3, 104.3, 43.6, 90°, 90°, 120°	104.9, 104.9, 42.7, 90°, 90°, 120°	104.9, 104.9, 42.7, 90°, 90°, 120°	105.3, 105.3, 42.9, 90°, 90°, 120°	137.9, 137.9, 172.8, 90°, 90°, 120°
Resolution range (Å)	34.1–2.0 (2.1–2.0)	45.4–1.8 (1.9–1.8)	45.4–2.4 (2.6–2.4)	52.6–2.5 (2.6–2.5)	43.6–3.3 (3.6–3.3)
Total reflections	171,250 (26,903)	261,119 (40,636)	130,070 (15,867)	73,821 (6,398)	131,292 (3,251)
Unique reflections	35,716 (5,693)	25,124 (4,007)	10,840 (1,719)	9,612 (1,042)	24,350 (650)
Completeness (%)	99.2 (98.0)	99.5 (98.8)	99.9 (99.4)	99.9 (99.5)	88.5 (47,8)
Redundancy	4.8 (4.7)	10.4 (10.1)	12.0 (9.2)	7.7 (6.1)	5.4 (5.0)
I/σ (I)	10.2 (2.0)	14.7 (2.6)	16.3 (2.0)	6.1 (1.8)	17.9 (1.8)
R-meas (%)	12.8 (92.1)	13.8 (97)	15.5 (126.5)	16.6 (55.4)	8.5 (71.5)
Refinement
R-work (%)		16.8	18.1	17.8	26.3
R-free (%)		21.0	25.2	22.9	29.0
Number of atoms
Protein		2,029	2,022	1,995	2,028
Nucleotide		—	27	31	92
Water		268	81	148	0
Average B-factor
Protein		21.5	40.2	21.6	92.1
Nucleotide		—	29.7	24.4	39.6
Water		33.1	38.8	22.5	0
RMSD
Bonds (Å)		0.007	0.007	0.009	0.013
Angles (°)		1.004	1.144	1.195	2.536
Ramachandran (%)
Favorable		98.4	95.6	98.0	85.1
Allowed		1.6	4.0	2.0	11.4
Outliers		0.0	0.4	0.0	3.5

Statistics for high-resolution shell are shown in parentheses.

Fig. S1.

Nucleotide binding and oligomerization of FleQ. (A) Crystallographic identification of ligands bound to the AAA+ domain of Pa_FleQ. Omit |F_o| − |F_c| maps are contoured at 3.0 σ (black). The observed densities are consistent with the identity of the nucleotides used in cocrystallization trials [ATP-γ-S (Right, Upper), ADP (Right, Lower), c-di-GMP (Left)]. (B) Negative stain electron microscopy analysis of full-length P. aeruginosa FleQ. A typical micrograph showing single particles (Left) and an autorotation function of a class-average particle highlighting the sixfold auto-rotation axis (Right) are shown. Nominal magnification, Left: 49,000×, boxed particle class average: 170 × 170 pixels with a pixel size of 2.17 Å. (C) Accessibility of the c-di-GMP binding site in the context of a ring-shape AAA+ domain hexamer. Side views of a Pa_FleQ^AAA hexameric ring reconstructed from crystal symmetry are shown. Residues involved in c-di-GMP binding are colored green. Subdomains SD1 and SD2 of one protomer are colored as in Fig. 1; the other symmetry related protomers are colored in shades of gray. σ⁵⁴ interaction loops L1 and L2 are presented in magenta and cyan, respectively. (Right) One of the protomers in a FleQ ring-shaped hexamer is shown in cartoon representation. The panel highlights that c-di-GMP-binding, as observed in the cocrystal structure, displaces the N-terminal tail of the AAA+ domain, which is incompatible with this ring-shaped oligomer form. (D) Role of Walker A residue K¹⁸⁰ in FleQ AAA+ P-loop and hexameric ring stability. Residue K¹⁸⁰ occupies a pivotal position for P-loop stabilization, interacting with the main-chain carbonyl oxygen of G¹⁷⁴ and E¹⁷⁵ and the side chain of T¹⁷⁸. Additionally, K¹⁸⁰ is involved in the interaction network of the ADP/ATP-γ-S β–phosphate located at the N-terminal of helix α7 (the sulfate anion shown here occupies the nucleotide’s β-phosphate position in the apo structure). Interactions between the side chains of S¹⁷⁶ and E¹⁸¹ with E³⁰¹ and R²⁶⁰ from neighboring protomers, respectively, contribute to hexameric ring formation.

Although the asymmetric units of all three structures contain a single polypeptide chain, a view along the sixfold symmetry axes reveals protomer packing similar to the hexameric rings of activated canonical AAA+ proteins (Fig. 1D) (28). Rather than forming discrete oligomers, however, the protomers arrange in continuous spirals along the screw axes (Fig. 1D, Right). Interestingly, similar spiral arrangements are observed in the crystal packing of the isolated AAA+ domain of Escherichia coli PspF (Ec_PspF^AAA), a bEBP known to form discrete hexameric rings in solution (29). PspF ring oligomers are believed to undergo sequential hydrolysis where asymmetry in the intersubunit interactions confers cooperativity in nucleotide binding and catalysis (30). In addition, solution structural studies on ring hexamers of another nonhelical bEBP, NtrC1, revealed that one of the intersubunit interfaces is significantly less tight, which led to a mechanistic model where bEBP planar ring split is necessary to drive interactions with the σ⁵⁴-RNAP holoenzyme and ATP-dependent remodeling (31).

Because the formation of ring-shaped FleQ assemblies would only require subtle changes in the protomer-packing interfaces observed in the crystalline spirals, we continued by examining the actual oligomerization states of FleQ in solution. To this end, we visualized full-length FleQ (Pa_FleQ^Full) by negative-stain electron microscopy following gel filtration in the presence of the transition-state analog ADP-vanadate. Single-particle analysis of the micrographs failed to reveal any filamentous structures but showed discrete hexameric species that are consistent with canonical bEBP ring assembly rather than helical organization (Fig. S1B). Although these experiments were performed under saturating conditions of a single, nonhydrolyzable nucleotide species and more studies are necessary to examine potential conformational changes and asymmetry along the hydrolytic cycle, it is possible that the crystal packing of the isolated AAA+ domain from both Pa_FleQ and Ec_PspF in the form of open oligomers reflects a conserved mechanism for bEBP activation and function.

Molecular Determinants for c-di-GMP Recognition by FleQ’s AAA+ Domain.

In the presence of c-di-GMP, a purified AAA–HTH dual domain construct of P. aeruginosa FleQ (Pa_FleQ^AAA-HTH, residues 137–477) crystallized in space group P6₃22 with a single protomer per asymmetric unit. Although the crystals were characterized by extremely high solvent content (∼80%) and suboptimal anisotropic diffraction, the structure was determined by Se-SAD to 3.3 Å resolution with good refinement statistics (Table S1). The refined electron density map covers residues 138–393 and excludes the C-terminal HTH domain, likely because of flexibility in the loosely packed crystals and insignificant contribution to the X-ray diffraction data.

Two molecules of c-di-GMP bind in their preferred intercalated dimer conformation in a region between subdomains SD1 and SD2 that is different from the ATP-binding active site (Fig. 2A and Fig. S1A). An earlier modeling study on a FleQ homolog from Vibrio cholerae, FlrA, suggested that residues corresponding to R¹⁴⁴ and R¹⁸⁵ of P. aeruginosa FleQ constitute a conserved c-di-GMP binding site at the REC and AAA+ domain interface (26), which was recently confirmed by another modeling-based study (32). The Pa_FleQ^AAA-HTH–c-di-GMP cocrystal structure not only confirms the importance of these residues for c-di-GMP recognition, but also reveals a composite binding site with contributions from three key motifs within a single FleQ protomer (Fig. 2A). Cis-acting residues include: a proximal c-di-GMP-binding switch located at the N-terminus of the AAA+ domain that is composed of the LFR¹⁴⁴S motif (R-switch); residues R¹⁸⁵ and N¹⁸⁶ in helix α7 of subdomain SD1 (post-Walker A); and finally, a distal c-di-GMP-binding motif comprising the ExxxR³³⁴ sequence in helix α13 of subdomain SD2 (Fig. 2 A and B). Isothermal titration calorimetry confirms both the stoichiometry of binding (∼2:1 for c-di-GMP:Pa_FleQ^Full with an apparent affinity constant K_d of ∼4.1 μM) and the role of conserved arginine residues in ligand coordination (Fig. S2 A and B).

Fig. 2.

Crystal structure of c-di-GMP–bound Pa_FleQ^AAA-HTH reveals major oligomeric rearrangements. (A) Close-up view of the c-di-GMP binding site. Interactions (green dashes) between intercalated c-di-GMP dimer and conserved dinucleotide binding residues R¹⁴⁴, R¹⁸⁵, N¹⁸⁶, E³³⁰, and R³³⁴ are shown. A crystal symmetry-related protomer that completes the c-di-GMP binding site is shown in gray. Labels that refer to the symmetric protomer contain asterisks. (B) Sequence conservation of the c-di-GMP binding site. Sequence logo representations of multisequence alignment for FleQ and NtrC1 homologs cover the three motifs involved in c-di-GMP recognition within the AAA+ domain. bEBPs sequences without a regulatory N-terminal domain or with REC, GAF or FleQ-like N-terminal domains were selected from the RP55 database with HMMER (60) and logos were generated with Skylign (61). Representative sequences are shown on top of the logos: P. aeruginosa FleQ (Pa_FleQ); V. cholerae FlrA (Vc_FlrA); X. citri FleQ (Xc_FleQ); V. cholerae VpsR (Vc_VpsR); A. aeolicus (Aa_NtrC1); Geobacter sulfurreducens NtrC (Gs_NtrC). (C) Comparison between c-di-GMP and ATP-γ-S–bound structures. Superposition of the respective SD1 domains reveals a 9° rigid body rotation (black arrow) of the SD2 subdomain relative to the SD1 domain upon c-di-GMP binding. The c-di-GMP–bound Pa_FleQ^AAA-HTH structure is colored as in Fig. 1; bound nucleotides are shown as sticks (yellow carbon atoms). (D) The N-terminal segment of FleQ’s AAA+ domain partially occupies the ATP site in the c-di-GMP–bound structure. Solid surface representation of S¹⁴⁵’s side chain and G¹⁷⁹’s main chain indicate steric hindrance at the ATP binding site caused by the conformational change induced by c-di-GMP. (E) Assembly composed of head-to-head dimers of c-di-GMP–stabilized trimers in two orthogonal views. Pa_FleQ^AAA-HTH crystal symmetry-related protomers are colored in different shades of gray. (F) Protein–protein interfaces within the c-di-GMP–bound Pa_FleQ^AAA-HTH dimer-of-trimers. The assembly involves head-to-head SD1/SD2 (Left) and SD2/SD2 (Center) interfaces and a c-di-GMP–mediated (Right) interface. Dissociation energies (ΔⁱG) and corresponding interaction-specificities (ΔⁱG P value) for each individual interface (green dashes) were calculated with the PISA server (36).

Fig. S2.

ITC of full-length FleQ (Pa_FleQ^Full and Xc_FleQ^Full) with c-di-GMP. Wild-type protein or mutants with site-directed alteration in c-di-GMP binding residues, which were identified based on the cocrystal structure, are shown. (A) Titration of c-di-GMP to wild-type Pa_FleQ^Full shows a stoichiometry of 2:1 (c-di-GMP:FleQ) with a K_d of 4.1 µM. (B) Mutations R¹⁸⁵E and R³³⁴E completely abolish ligand binding, as evidenced by the endothermic peak profiles that are similar to those observed by c-di-GMP dilution into buffer (Upper). Wild-type Pa_FleQ^Full data were reproduced in Fig. 4 for comparison with other site-directed mutant proteins. (C) Titration of c-di-GMP to wild-type, full-length Xanthomonas citri FleQ (Xc_FleQ^Full).

A systematic sequence analysis of bEBPs revealed a higher degree of conservation of c-di-GMP binding motifs (R-switch, post-Walker A and ExxxR consensus motifs) within FleQ homologs (Fig. 2B). This group includes the c-di-GMP–responsive regulator FlrA from V. cholerae, for which the importance of R¹³⁵ and R¹⁷⁶ (corresponding to FleQ R¹⁴⁴ and R¹⁸⁵, respectively) for c-di-GMP binding has been determined experimentally (26). In particular, the coconservation of all three c-di-GMP-binding motifs, and especially their arginine residues, is striking (Fig. 2B). Consistent with this notion, isothermal titration calorimetry (ITC) experiments with a FleQ homolog from Xanthomonas citri (XAC1967), which contains divergent c-di-GMP binding motifs (Fig. 2B), failed to show c-di-GMP binding (Fig. S2C). On the other hand, co-occurring arginine residues important for c-di-GMP binding were found in several bEBPs with different N-terminal regulatory domains (Fig. 2B, Table 1, Fig. S3, and Table S2). Of 4,632 bEBPs mined from a representative proteome database (33), ∼3% presented conserved c-di-GMP binding residues, suggesting that c-di-GMP regulation of bEBP-mediated transcription may be more widely used than previously anticipated, potentially participating in other cellular processes beyond flagellar regulation and EPS secretion (Fig. 2B and Table 1). Additionally, the absence of the aforementioned conserved c-di-GMP binding motifs in the atypical bEBP VpsR from V. cholerae (34) and in AAA+ ATPases involved in multiple bacterial secretion systems (35), all shown to bind to c-di-GMP, further suggest that bacterial AAA+ domains may represent ubiquitous targets for c-di-GMP, yet with distinct binding sites.

Table 1.

Sequences of bEBPs presenting conserved c-di-GMP binding motifs

N-terminal regulatory domain	Total of sequences in RP55 database*	Sequences with conserved c-di-GMP binding motifs†
FleQ‡	90	67 (74.4%)
GAF	238	5 (2.1%)
—	1257	8 (0.6%)
REC	3047	62 (2.0%)

^*Sequences of FleQ and NtrC (REC) homologs used for the Logo generation in Fig. 2B.

^†All three motifs conserved; see Table S2 for accession numbers.

^‡FleQ refers to the N-terminal degenerated REC domain.

Fig. S3.

Alignment of representative bEBPs with putative c-di-GMP binding site. The alignment was generated with representative sequences from Table 1 and Table S2. Coloring of conserved features (Walker A, Walker B, Loop L1, Loop L2, and R finger) follows the same scheme as introduced in Fig. 1A. C-di-GMP binding motifs are highlighted in red boxes, conserved residues (compared with Pa_FleQ) in these motifs are in bold. Red arrowheads below the alignment point to conserved arginine (or lysine) residues that coincide with arginine residues in FleQ important for c-di-GMP binding. Black arrowheads mark other c-di-GMP–interacting residues identified in the structure. Dashed lines separate sequences by nature of their N-terminal domain as in Table 1 [FleQ-like, REC, GAF, and without recognizable N-terminal domains (-), indicated in the right column]. The first and second column list Uniprot identifiers and residue numbering, respectively. The following sequences were used: A8H297_SHEPA: Shewanella pealeana (strain ATCC 700345/ANG-SQ1); Q15RD1_PSEA6: Pseudoalteromonas atlantica (strain T6c/ATCC BAA-1087); A0A0C5WCX2_9GAMM: Photobacterium gaetbulicola Gung47; Q9KQ66_VIBCH: Vibrio cholerae serotype O1 (strain ATCC 39315/El Tor Inaba N16961); Q3J8M6_NITOC: Nitrosococcus oceani (strain ATCC 19707/NCIMB 11848); Q21IM5_SACD2: Aaccharophagus degradans (strain 2–40/ATCC 43961/DSM 17024); Q74DJ6_GEOSL: Geobacter sulfurreducens (strain ATCC 51573/DSM 12127/PCA); B9M0Q9_GEODF: Geobacter daltonii (strain DSM 22248/JCM 15807/FRC-32); A0R7M9_PELPD: Pelobacter propionicus (strain DSM 2379); S6BIE6_9GAMM: endosymbiont of unidentified scaly snail isolate Monju; D0KXW5_HALNC: Halothiobacillus neapolitanus (strain ATCC 23641/c2) (Thiobacillus neapolitanus); A0A075WZ96_9BACT: Thermodesulfobacterium commune DSM 2178.

Table S2.

bEBPs predicted to contain a conserved c-di-GMP binding motif (sequences associated with Table 1, column 3)

FleQ domain		GAF domain		No N-terminal domain		REC domain
Accession	Organism	Accession	Organism	Accession	Organism	Accession	Organism
Q0AAS7	Alkalilimnicola ehrlichii	Q6NA12	Rhodopseudomonas palustris	S6BIE6	unclassified	B5EAE4	Geobacter bemidjiensis
Q8ECB1	Shewanella oneidensis	H9U.K.K9	Spirochaeta africana	W0TKX7	Thiolapillus brandeum	E6VV23	Desulfovibrio aespoeensis
K4KHP1	Simiduia agarivorans	F8F1U5	Treponema caldarium	D0KXW5	Halothiobacillus neapolitanus	F6B7S3	Desulfotomaculum carboxydivorans
A0A0C5WCX2	Photobacterium gaetbulicola	A0A075WZ96	Thermodesulfobacterium commune	A4A600	Congregibacter litoralis	D3FVN4	Bacillus pseudofirmus
Q9KQ66	Vibrio cholerae	B3DZA5	Methylacidiphilum infernorum	Q31FN5	Thiomicrospira crunogena	I8RAD9	Pelosinus fermentans
W8L6B7	Halorhodospira halochloris			D3RRD4	Allochromatium vinosum	C6BU16	Desulfovibrio salexigens
D4ZIA0	Shewanella violacea			I3Y607	Thiocystis violascens	F8A837	Thermodesulfatator indicus
A3QCN3	Shewanella loihica			D3S9F9	Thioalkalivibrio sp.	F2NE61	Desulfobacca acetoxidans
Q12PM0	Shewanella denitrificans					B5EBB4	Geobacter bemidjiensis
B1KLH5	Shewanella woodyi					A5GDQ0	Geobacter uraniireducens
B8GT39	Thioalkalivibrio sulfidiphilus					C8W398	Desulfotomaculum acetoxidans
F2G3X6	Alteromonas macleodii					A0A075WYK6	Thermodesulfobacterium commune
A1WUC6	Halorhodospira halophila					A7IE40	Xanthobacter autotrophicus
Q7MIN1	Vibrio vulnificus					E1WYR5	Halobacteriovorax marinus
Q085N4	Shewanella frigidimarina					F6CNR7	Desulfotomaculum kuznetsovii
Q3J8M6	Nitrosococcus oceani					A4J4N7	Desulfotomaculum reducens
Q21IM5	Saccharophagus degradans					A4J616	Desulfotomaculum reducens
E1VPK4	Gamma proteobacterium HdN1					F8ABW9	Thermodesulfatator indicus
C1DQF1	Azotobacter vinelandii					B8FJT8	Desulfatibacillum alkenivorans
C5BRZ8	Teredinibacter turnerae					Q6MHV3	Bdellovibrio bacteriovorus
B3PEX8	Cellvibrio japonicus					D8PFF7	Nitrospira defluvii
A0A0C5VAJ5	Gynuella sunshinyii					W5X1M1	Bdellovibrio bacteriovorus
R4YT01	Oleispira antarctica					A8MGT2	Alkaliphilus oremlandii
U4K2L0	Vibrio nigripulchritudo					F6B3K1	Desulfotomaculum carboxydivorans
A0A098G2Y6	Legionella fallonii					M4VEB7	Bdellovibrio exovorus
F7YNB5	Vibrio anguillarum					F6DU39	Desulfotomaculum ruminis
Q87MJ6	Vibrio parahaemolyticus					Q39XH0	Geobacter metallireducens
F0LV23	Vibrio furnissii					C8WYS9	Desulfohalobium retbaense
I1XLA6	Methylophaga nitratireducenticrescens					Q3A223	Pelobacter carbinolicus
D3HJ18	Legionella longbeachae					Q2IMP6	Anaeromyxobacter dehalogenans
Q5ZX79	Legionella pneumophila					I0GJJ1	Caldisericum exile
G3XCV0	Pseudomonas aeruginosa					A0A075K6I7	Pelosinus sp.
K0WPZ8	Pseudomonas fluorescens					Q3B5Z9	Chlorobium luteolum
A0A089ZPQ9	Pseudomonas rhizosphaerae					I4D1F0	Desulfosporosinus acidiphilus
L0GTK0	Thioflavicoccus mobilis					Q2IHE3	Anaeromyxobacter dehalogenans
Q4KG73	Pseudomonas fluorescens					Q39XK6	Geobacter metallireducens
Q485M5	Colwellia psychrerythraea					C7PEX2	Chitinophaga pinensis
A0A077LJ97	Pseudomonas sp.					Q74DJ6	Geobacter sulfurreducens
K0C4W0	Cycloclasticus sp.					Q1MQI6	Lawsonia intracellularis
G4QLX6	Glaciecola nitratireducens					A7HIM1	Anaeromyxobacter sp.
A0A089WSC4	Pseudomonas cremoricolorata					M1WKA5	Desulfovibrio piezophilus
G8Q0L8	Pseudomonas fluorescens					Q3A356	Pelobacter carbinolicus
E1SUP9	Ferrimonas balearica					B9M0Q9	Geobacter daltonii
A0A097QM59	Vibrio coralliilyticus					J7IP96	Desulfosporosinus meridiei
A4VMJ3	Pseudomonas stutzeri					A0R7M9	Pelobacter propionicus
Q884Y2	Pseudomonas syringae					H1XQY8	Caldithrix abyssi
A1U263	Marinobacter hydrocarbonoclasticus					A6TT75	Alkaliphilus metalliredigens
W8RVQ1	Pseudomonas stutzeri					Q39VU3	Geobacter metallireducens
Q2SBU8	Hahella chejuensis					F3Z0Y5	Desulfovibrio africanus
A0A098GFT6	Tatlockia micdadei					D0LFQ8	Haliangium ochraceum
I1YKN4	Methylophaga frappieri					C8W392	Desulfotomaculum acetoxidans
D5BV80	Nitrosococcus halophilus					E8RDW7	Desulfobulbus propionicus
F9T4M1	Vibrio tubiashii					D6Z2J3	Desulfurivibrio alkaliphilus
B7VL01	Vibrio tasmaniensis					F0JC51	Desulfovibrio desulfuricans
Q6LTP6	Photobacterium profundum					G7W595	Desulfosporosinus orientis
Q5E3P5	Vibrio fischeri					Q2LSK2	Syntrophus aciditrophicus
A0A0A8UN75	Legionella hackeliae					D3PA19	Deferribacter desulfuricans
A1S800	Shewanella amazonensis					B3E969	Geobacter lovleyi
M5E625	Thalassolituus oleivorans					A0LPT2	Syntrophobacter fumaroxidans
A8H297	Shewanella pealeana					D5MM35	Candidatus Methylomirabilis oxyfera
Q15RD1	Pseudoalteromonas atlantica					I0AG67	Ignavibacterium album
A0A024HES6	Pseudomonas knackmussii					A5GFK3	Geobacter uraniireducens
Q88ET0	Pseudomonas putida
F6A8I2	Pseudomonas fulva
S6AU94	Pseudomonas resinovorans
A4XW68	Pseudomonas mendocina
C3K0U9	Pseudomonas fluorescens

Consequences of c-di-GMP Binding on FleQ’s Tertiary and Quaternary Structure.

Compared with apo, ADP-, and ATP-γ-S–bound Pa_FleQ^AAA structures, the conformation of the c-di-GMP-bound AAA+ ATPase σ⁵⁴-interaction domain reveals novel and unexpected features at both the tertiary and quaternary structure levels. Protomer superposition using subdomain SD1 as the reference reveals a 9° rigid body movement of subdomain SD2 to a more compact, c-di-GMP–stabilized conformation relative to the apo-, ADP-, and ATP-γ-S–bound structures (Fig. 2C). It has been proposed previously that c-di-GMP competes with ATP for binding to the AAA+ ATPase domain active site based on the fact that c-di-GMP competitively inhibits ATP hydrolysis and that a Walker A disruptive mutation (FleQ K¹⁸⁰A) abolishes c-di-GMP binding to FleQ (21). However, this model is at first sight inconsistent with the crystallographic analysis. Superposition of the c-di-GMP– and ATP-γ-S–bound Pa_FleQ^AAA structures reconciles the apparent contradictions by revealing that c-di-GMP binding shifts the N terminus of the AAA+ domain into the ATP-binding pocket (Fig. 2D), thus acting as an allosteric inhibitor of the protein’s hydrolytic activity and, by extension, of σ⁵⁴-dependent transcription activation. Importantly, in the context of the full-length protein, such an N-terminal movement in the FleQ AAA+ domain would cause a drastic repositioning of the preceding FleQ^REC module, possibly contributing to the disruption of the active-state hexameric ring (Fig. S1C). Finally, residue K¹⁸⁰ in the Walker A motif occupies a pivotal location between the two AAA+ subdomains and plays a significant role in stabilizing interprotomer interaction within Pa_FleQ^AAA hexameric rings (Fig. S1D). A mutation in this position could interfere with both c-di-GMP–mediated AAA+ interdomain movement and oligomeric assembly.

The structural organization of the c-di-GMP binding pocket provides further evidence for negative coupling of this site with the ATPase active site. The highly conserved residues R¹⁴⁴, R¹⁸⁵, and R³³⁴ are not only directly involved in c-di-GMP coordination (Fig. 2A and Fig. S2), but also directly or indirectly stabilize ATP binding, as observed in the ADP and ATP-γ-S–complexed Pa_FleQ^AAA crystal structures (Fig. 1C). In particular, residue R³³⁴ is in direct contact with the ribose moiety of ATP, while also stabilizing the N-terminal AAA+ domain segment through π-stacking interactions with R¹⁴⁴ (Fig. 1C). Similarly, residue R¹⁸⁵ forms a salt bridge with the Walker A residue E¹⁸¹, and might also sense nucleotide binding (Fig. 1C). All these specific interactions important for ATP binding and hydrolysis are disrupted in the c-di-GMP–bound state.

The most striking feature of the c-di-GMP–bound structure, however, is the profound reorganization of crystallographic interfaces, which suggests dramatic quaternary structure reorganization upon allosteric second-messenger recognition. The observed c-di-GMP–mediated subdomain movement and N-terminal restructuring appear incompatible with the maintenance of canonical bEBP ring hexamers. Indeed, symmetry-related protomers arrange into a distinct hexameric species where c-di-GMP–stabilized trimers dimerize in a head-to-head fashion (“dimer-of-trimers”) (Fig. 2E). Dissociation energy analyses of protein-protein and c-di-GMP–protein interfaces (SD1/SD2, SD2/SD2, and c-di-GMP–mediated interfaces) suggest biological relevance for this unique oligomeric arrangement (Fig. 2F) (36). Within the trimeric assembly mediated by c-di-GMP, a total buried surface area of about 1,400 Ų yields a gain of −14.4 kcal/mol in solvation energy. The hexamer formed by SD1/SD2 and SD2/SD2 subdomain pairings further occludes ∼4,300 Ų of surface area, with a total gain in solvation energy of −23.7 kcal/mol. To our knowledge, such architecture is unprecedented for bEBP family members and underscores the unique mechanism of FleQ regulation and function.

C-di-GMP Drives FleQ Oligomeric Reorganization in Solution.

External signal stimuli, such as ligand binding (ATP or promoter DNA) and protein phosphorylation at the REC domain, usually trigger the switch of inactive bEBP dimers to ring-like activated hexamers (17, 19, 37). In this regard, FleQ behavior deviates once again from canonical bEBPs, by being able to spontaneously hexamerize in solution (21). In addition, although the N-terminal domain of FleQ shares sequence homology with canonical REC domains, it lacks both sequence conservation at the putative phosphorylation site and a cognate histidine kinase (12), making regulatory phosphorylation unlikely.

To evaluate the role of ligand binding on FleQ oligomerization in solution, we captured discrete oligomeric species through chemical cross-linking with the amine-reactive compound ethylene glycol bis-(succinimidyl succinate) (EGS). In agreement with previous results (21), wild-type full-length FleQ (Pa_FleQ^Full) showed a dimeric fraction as well as discrete higher-order species (mostly tetramers and hexamers), regardless of the absence or presence of ATP-γ-S (Fig. 3). A cross-linking time-course revealed a gradual accumulation of the higher oligomeric species, suggesting an equilibrium between FleQ oligomers in solution (Fig. 3A). Most notably, however, addition of c-di-GMP stalled FleQ oligomerization beyond a dimeric species, regardless of ATP-γ-S availability (Fig. 3B). The striking effect of c-di-GMP on FleQ’s oligomeric assembly mirrors the dinucleotide’s effect on FleQ’s protease sensitivity (21), and corroborates c-di-GMP–driven quaternary structure changes inferred from the crystal structures (Figs. 1 and 2).

Fig. 3.

Cross-linking analysis of full-length FleQ oligomerization in solution. SDS/PAGE analyses of PA_FleQ^Full protein (10 μM) incubated with or without the primary amine-specific cross-linker EGS are shown. (A) FleQ cross-linking time course in the absence of nucleotides. The sample in lane 2 was not incubated with EGS. All other samples were incubated with EGS (250 µM). (B) Effects of nucleotides on FleQ cross-linking. Addition of nucleotides (100 μM) is indicated above the gel. Where indicated, samples were incubated for 30 min at room temperature in the presence of EGS (1 mM). Migration pattern of probable oligomeric species are indicated on the right of each gel.

Similar cross-linking experiments using FleQ constructs lacking the REC domain, Pa_FleQ^AAA and Pa_FleQ^AAA-HTH, showed less pronounced effects (Fig. S4). In particular, oligomerization of Pa_FleQ^AAA was severely compromised, with the protein migrating mostly as a monomer, independent of nucleotide (Fig. S4A). This result is consistent with a recent report that dimerization of the noncanonical REC domains is critical for FleQ oligomerization and function (32). Inclusion of the C-terminal HTH domain restored in-solution dimerization partially (Fig. S4B). Addition of c-di-GMP stablized higher-order oligomers of Pa_FleQ^AAA (and to a lesser extent Pa_FleQ^AAA-HTH) including trimers and hexamers, consistent with a change in quartenary structure and our crystallographic analyses. These results mirror studies on trunctation mutants of other bEBPs, which revealed oligomerization properties and functional activities distinct from their full-length counterparts. For example, removal of the N-terminal regulatory domain of Salmonella enterica serovar Typhimurium NtrC and Aquifex aeolicus NtrC1 inhibits or promotes, respectively, the formation of activated oligomeric ring assemblies (19, 20).

Fig. S4.

Cross-linking analysis of Pa_FleQ^AAA and Pa_FleQ^AAA-HTH oligomerization in solution. SDS/PAGE of Pa_FleQ^AAA (A) and Pa_FleQ^AAA-HTH (B) proteins (10 μM) incubated with the primary amine-specific cross-linker EGS (1 mM) for 30 min at room temperature are shown. Addition of nucleotides (100 μM) is indicated above the gel. Calculated monomer molecular weights: Pa_FleQ^AAA, 29.1 kDa; Pa_FleQ^AAA-HTH, 38.2 kDa. Migration pattern of probable oligomeric species are indicated on the right of each gel. Diagrams above each gel depict the domain architecture of the constructs used in each experiment.

Taken together, our data indicate that c-di-GMP antagonizes different FleQ oligomeric assemblies, in accordance with our structural analyzes. The cross-linking studies not only indicate a crucial role of the REC domain for spontaneous FleQ oligomerization but also suggest that hexamer formation may be a prerequisite of c-di-GMP binding.

Functional Analysis of Interface-Disruptive FleQ Variants Confirm Biological Significance of Crystallographic Oligomers.

We next designed a set of single-point mutations to confirm the role of crystallographic protein–protein and protein–ligand interfaces on full-length FleQ oligomerization, as well as on ATPase activity and c-di-GMP binding. Mutant I³⁷⁴E was designed to perturb formation of bEBP canonical ring-like hexamers apparent in the apo-, ADP-, and ATP-γ-S–bound Pa_FleQ^AAA structures; mutants R¹⁸⁵E, N¹⁸⁶A, E³³⁰A and R³³⁴E were designed to directly inhibit c-di-GMP binding; and two additional mutations were chosen to disrupt the SD1/SD2 (T¹⁴⁹E) or SD2/SD2 (V³⁸⁰E) interfaces observed in the c-di-GMP–bound Pa_FleQ^AAA-HTH structure (Fig. 4A). The proper folding and preserved stability of all mutants introduced in the full-length protein were verified by circular dichroism experiments of thermal denaturation (Fig. S5A).

Fig. 4.

Functional analyses of structure-guided Pa_FleQ^Full mutants. (A) Overview of site-directed mutants and predicted outcome. The diagram shows the position of each mutant on a cartoon of the hexameric ring or the dimer-of-trimers assembly. Predicted outcomes on FleQ function are indicated, distinguishing between indifferent (=), positive (↑), and negative (↓). (B) Cross-linking analysis. Mutant proteins were treated as described in Fig. 3, which shows the corresponding data for wild-type protein. (C) ATPase activity. Initial velocity of Pa_FleQ^Full variant-catalyzed ATP hydrolysis is shown expressed as percentage of wild-type Pa_FleQ^Full activity (1 μM protein; 1 mM ATP). The error bars represent SDs of triplicates. (D) c-di-GMP effect on ATP hydrolysis. ATPase activity of selected FleQ variants preincubated with c-di-GMP (50 μM). The error bars represent SDs of triplicates. (E) c-di-GMP binding. ITC of Pa_FleQ^Full T¹⁴⁹E and I³⁷⁴E. The heat change upon ligand binding (μcal/s) and fitted curves to the integrated heat peaks are shown in the top and bottom titration curves, respectively. Binding stoichiometry and apparent dissociation constants are indicated in the table inset.

Fig. S5.

Thermal stability and cross-linking analyses of Pa_FleQ^Full variants. (A) Protein stability. Thermal unfolding of full-length FleQ variants was monitored by plotting the intensity of the circular dichroism signal at 222 nm upon heat denaturation. Curves are plotted as fraction of completely denatured protein (90 °C). Ellipticity (θ) scans over a wavelength range of 200–260 nm at 20 °C are shown for all samples (Right). (B) Cross-linking analyses of Pa_FleQ^Full mutant variants. The impact of c-di-GMP (50 µM; Left) and increasing concentration of ATP-γ-S (10–100 µM; Right) on the oligomerization of Pa_FleQ^Full mutants (50 µM) was monitored by EGS-mediated cross-linking. Reactions were carried out in reaction buffer containing 250 µM EGS at room temperature for 30 min.

In the absence of c-di-GMP, higher molecular weight oligomers beyond dimers were severely disrupted only with the I³⁷⁴E mutant, consistent with our structure-based prediction (Fig. 4B). Mutants with disrupted c-di-GMP binding pocket (R¹⁸⁵E and R³³⁴E) retained wild-type–like oligomerization patterns, but were impaired in c-di-GMP’s antagonistic effect on assembly (Fig. 4B and Fig. S5B). Mutations designed to disrupt the interfaces SD1/SD2 and SD2/SD2 observed in the c-di-GMP–bound structure, T¹⁴⁹E and V³⁸⁰E, respectively, maintained FleQ’s propensity to form hexamers, rendering the protein apparently insensititve to c-di-GMP (Fig. 4B, lanes 5–6, and Fig. S4B). Hence, destabilization of the crystallographically observed dimer-of-trimers appears to have a positive effect on the stability of the canonical bEBP hexamer. Similar to wild-type FleQ, ATP-γ-S did not alter the oligomerization propensities of any of the mutants (Fig. S5B).

Based on the fact that FleQ is an active ATPase that is primarily dimeric in solution, it has been proposed previously that the dimers are the catalytically active conformation (21). As we show here, however, FleQ samples higher-order oligomeric conformations in solution, including hexameric rings (Fig. 3 and Fig. S1B), and its AAA+ domain crystallizes in a conformation consistent with the active-state ring hexamers of canonical bEBPs (Fig. 1D). These data prompted us to dissect the role of oligomerization on the protein’s catalytic activity. In ATPase activity assays, we observed a severe reduction in the ATP hydrolysis rate with the I³⁷⁴E mutant (Fig. 4C). Because the cross-linking indicates a defect in hexamer formation while dimerization is preserved in this mutant, these data assert the hexamer as the catalytically active conformation.

The ATPase activity data also confirm conformational coupling between the c-di-GMP– and ATP-binding pockets: disruption of the conserved R¹⁸⁵ and R³³⁴ residues abolishes not only c-di-GMP binding, but also ATPase activity (Fig. 4C and Fig. S2B), in line with their structural role in active-site stabilization (Fig. 1C). Mutants designed to disrupt crystallographic, c-di-GMP–mediated dimer-of-trimers, T¹⁴⁹E and V³⁸⁰E, have preserved and even enhanced ATPase activity (7.5- and 1.5-fold increase over wild-type for T¹⁴⁹E and V³⁸⁰E, respectively) (Fig. 4C). Analysis of the known inhibitory effect of c-di-GMP on FleQ’s ATPase activity (21) revealed that although the V³⁸⁰E mutation reduces c-di-GMP sensitivity only slightly, the high activity of the T¹⁴⁹E mutant remains unaffected (Fig. 4D).

To evaluate the impact of SD1/SD2 or canonical ring interface perturbations on c-di-GMP binding, we performed ITC experiments with FleQ variants T¹⁴⁹E and I³⁷⁴E, respectively. Under the same experimental conditions used for wild-type Pa_FleQ^Full, ITC results with both mutants revealed a severe reduction of heat released during the titration (Fig. 4E), indicating that disruptive mutations at both oligomeric interfaces observed in the distinct crystallographic hexamers (ring-shaped or dimer-of-trimers) (Figs. 1D and 2E) negatively affects c-di-GMP binding.

FleQ Oligomeric Reorganization Triggered by c-di-GMP Is Compatible with Its Repressor-to-Activator Functional Transition on EPS Gene Expression.

FleQ recognizes two sites (FleQ boxes 1 and 2), separated by 34 bp, in the pelA promoter region (Fig. S6) (13). Previous reports showed that FleQ, stabilized by FleN and ATP/ADP, binds to both sites simultaneously, promoting a kink in the DNA and thus repressing gene expression (13). C-di-GMP binding to FleQ relieves the DNA distortion and activates gene expression without affecting the apparent protein occupancy on DNA (13, 14). In a mutant strain with a fleN deletion, FleQ still acts as an ATP-independent repressor of pel expression and increased c-di-GMP levels induce FleQ-mediated pel expression, albeit at a reduced level (14). Hence, the ΔfleN genetic background offers a unique possibility to examine directly the role of FleQ’s quaternary structure on pel transcriptional repression and activation. To evaluate the physiological role of the structural transition observed in vitro on transcriptional regulation via FleQ, we tested the effect of the various FleQ variants on pel operon expression in vivo.

Fig. S6.

Model of full-length Pa_FleQ hexameric rings bound to pel promoter DNA. A full-length FleQ model was generated by combining the crystal structure of a Pa_FleQ^AAA hexameric ring with the REC and HTH domain models based on PDB ID codes 1NY5 and 4L5E, respectively (20, 62). The position of the REC and HTH domains were guided by the position of the N- and C-termini of FleQ^AAA protomers. A stretch of DNA with FleQ boxes 1 and 2 was modeled as well, with the sequence of the pel promoter shown below. The asterisk indicates the position of the hypersensitive site identified by DNase I footprinting experiments (13).

Wild-type or mutant FleQ was expressed from a single-copy under its native promoter in a ΔfleN background. Elevated c-di-GMP levels were achieved by the coexpression of the active diguanylate cyclase YfiN/TpbB/PA1120 (38, 39). In line with previous results, pel expression was markedly enhanced upon YfiN coexpression with wild-type FleQ, indicating c-di-GMP–dependent transcription derepression and activation (Fig. 5A) (14). As expected, both promoter repression and YfiN effects were abolished in a fleQ deletion strain. Consistent with the structural and biochemical data, c-di-GMP binding mutants (R¹⁸⁵E, N¹⁸⁶A, E³³⁰A, and R³³⁴E) exhibited markedly repressed pel transcription similar to wild-type fleQ, but remained virtually unchanged upon YfiN overexpression (Fig. 5A). The mutant T¹⁴⁹E, predicted to disrupt the c-di-GMP–stablized crystallographic dimer-of-trimers, also yielded repressed pel transcription, independent of c-di-GMP. Consistent with its residual c-di-GMP sensitivity as demonstrated by the ATPase data, mutant V³⁸⁰E showed a somewhat derepressed pel transcription upon coexpression of a diguanylate cyclase (Figs. 4 and 5A). Notably, the I³⁷⁴E mutant was characterized by constitutively active pel expression, suggesting that hexameric ring formation is required for repression of the operon (Fig. 5A).

Fig. 5.

Effect of FleQ site-directed mutants on pel transcriptional control. (A) Expression data. Relative pelA transcript levels in strains of P. aeruginosa PAO1 that lack fleN (designated WT) or fleN and fleQ (designated/ΔfleQ) were quantified by RT-PCR. Structure-guided mutations in the fleQ gene were expressed from a single copy introduced into the ΔfleN/ΔfleQ genome. Transcripts were measured with (blue; + c-di-GMP) or without (gray; − c-di-GMP) expression of the diguanylate cyclase PA1120 from a plasmid. Unbroken (ΔfleN/ΔfleQ) and dashed (ΔfleN; wild-type FleQ) colored lines indicate reference pel expression levels at low and high c-di-GMP conditions. (B) Molecular model of FleQ-mediated pel transcription. The diagram depicts a model incorporating our structural and functional data, focusing on the events in the absence of FleN and ATP.

Taken together, the expression data in mutant genetic backgrounds corroborate the structural and biochemical results, indicating an equilibrium between ATPase-active, pel repressive hexamers, and other less-functional oligomeric species. However, ATPase activity is not required for FleQ’s function as a transcriptional repressor at the pel operon. C-di-GMP appears to exert its stimulatory effect on pel expression on FleQ variants that retain their ability to form hexamers and display an intact dinucleotide binding site (e.g., wild-type and V³⁸⁰E).

Discussion

In response to specific stimuli, canonical bEBPs from the large AAA+ family of σ⁵⁴-dependent transcription regulators couple ATP hydrolysis to promoter-RNAP initiation complex remodeling to control associated gene expression (15). Although unusual, some bacterial activators, such as Pseudomonas putida PhhR, E. coli TyrR, Rhodobacter capsulatus HupR, and Myxococcus xanthus HsfA share similar domain architecture with bEBPs, but activate the σ⁷⁰-RNAP holoenzyme instead of the σ⁵⁴-associated transcriptional machinery (40–42). P. aeruginosa FleQ combines both canonical bEBP σ⁵⁴- and atypical σ⁷⁰-dependent transcriptional mechanisms to inversely control flagellar motility and EPS production, respectively (13, 14, 21). Significantly, FleQ is also regulated by the bacterial second messenger dinucleotide, c-di-GMP (14). To gain molecular insight into the mechanism of c-di-GMP–dependent regulation of P. aeruginosa EPS gene expression by FleQ, we determined structures of its AAA+ domain in the apo-state and bound to ADP, ATP-γ-S, and c-di-GMP. Analyses of these structures, combined with in vitro and in vivo functional studies, revealed that P. aeruginosa FleQ undergoes a major conformational change upon c-di-GMP binding, concomitant with oligomeric rearrangement.

On one hand, the results presented here indicate a hexameric FleQ as the functional pel transcription repressor and as the target of c-di-GMP, because destablilization of the canonical AAA+ ring interfaces reduces c-di-GMP binding to FleQ (mutant I³⁷⁴E) (Figs. 4 and 5 and Fig. S1C). On the other hand, perturbations at the unique interfaces that support the crystallographic, c-di-GMP–bound FleQ dimer-of-trimers (e.g., mutants T¹⁴⁹E and V³⁸⁰E) not only render the protein variants insensitive to c-di-GMP, but also stabilize highly ATPase-active FleQ oligomers (Fig. 4). This finding may suggest that different FleQ oligomers with distinct properties exist in equilibrium (Fig. 4).

In light of our results and previous studies on the regulation of EPS gene expression, we propose that canonical bEBP ring-like FleQ hexamers (Fig. 1D and Fig. S1B) bind both boxes 1 and 2 at the pelA promoter, which inhibits transcription initiation (Fig. 5B). Apparently, FleQ’s function as a pel repressor in the absence of the cofactor FleN does not require its ATPase activity, but depends on the protein’s ability to form higher-order oligomers, in particular hexameric rings (Figs. 4 and 5). The hexamer-to-DNA interaction model is also compatible with the promoter bending observed previously (Fig. S6) (13). The current hypothesis is that association of ATP-bound FleN to the FleQ:DNA complex is necessary to increase DNA curvature in vitro and that the FleQ–FleN interaction cooperatively enhances c-di-GMP binding (13, 21), consistent with full activation of FleQ-c-di-GMP–mediated pel transcription requiring FleN (14). In analogy, the repression of σ⁷⁰-dependent tyrR expression by the atypical bEBP TyrR from E. coli occurs via binding of tyrosine-stabilized hexamers to two adjacent sites at the promoter DNA (14, 43–46), suggesting that the mechanism described here might be a recurrent form of transcriptional repression by atypical bEBPs.

In the presence of c-di-GMP, FleQ undergoes conformational changes and alters its quaternary structure (Figs. 2 and 3), which may decrease affinity for at least some promoter interaction sites, thus relieving DNA bending and activating gene transcription (Fig. 5 and Fig. S6) (14). bEBPs interact with σ⁵⁴ via two loops, L1 and L2. However, in our structures, these loops are accessible in the c-di-GMP–stabilized oligomer and not subject to conformational changes (Fig. 2C), suggesting a mode of transcriptional activation different from canonical bEBPs (18). The lack of conservation or absence of σ⁵⁴-interacting loops in other σ⁷⁰-dependent bEBP, such as E. coli TyrR and P. putida PhhR, respectively, argues further against their role in the direct activation of pel expression (40, 46).

Our data explain the molecular mechanism underlying c-di-GMP regulation of FleQ’s functions as repressor and activator of pel expression, which likely depends on σ⁷⁰ and is only moderately affected by the cofactor ATPase FleN. Future studies will focus on the roles of the N-terminal regulatory domains upstream of the AAA+ core and accessory proteins such as FleN in the context of FleQ’s oligomeric states, and in particular on the regulation of flagellar gene regulation. The latter process requires FleQ’s ATPase activity and σ⁵⁴, distinct from the mechanism described here, and presumably mirrors those of canonical bEBPs. How σ⁵⁴- and ATPase-dependent processes are coordinated with σ⁷⁰-dependent but ATPase-independent regulation at different promoters remains an open question.

Materials and Methods

Recombinant DNA Techniques.

DNA manipulations were carried out using standard molecular biology techniques. For protein expression and purification, the ORFs of full-length and truncated FleQ variants were PCR-amplified from P. aeuruginosa PA14 genomic DNA and cloned in-frame into a modified pET28a expression plasmid (Novagen). Expression yields recombinant proteins with an N-terminal hexahistidine-SUMO moiety that can be liberated by incubation with yeast Ulp1 protease. FleQ point mutants were generated using QuikChange site-directed mutagenesis kit (Agilent), following the manufacturer’s instructions. Successful mutagenesis was confirmed by DNA sequencing.

Protein Expression and Purification.

Native and selenomethionine-derivatized proteins were overexpressed in E. coli BL21 (DE3) (Novagen) or T7 Crystal Express (New England Biolabs) cells, respectively. For the expression of native proteins, cells were grown at 37 °C in lysogeny broth (LB) medium supplemented with 50 μg/mL kanamycin. At a cell density corresponding to an absorbance of 0.6 at 600 nm (OD₆₀₀), the temperature was reduced to 20 °C and protein production was induced with 1 mM IPTG. Selenomethionine-derivatized proteins were produced in cells grown in M9 minimal medium supplemented with 50 μg/mL kanamycin, 1 μg/mL thiamine, 1 μg/mL biotin, 0.4% glucose, and 40 μg/mL of each of the 20 amino acids with selenomethionine substituting for methionine. Protein expression was induced at an OD₆₀₀ of 0.4. After overnight expression, the cells were harvested by centrifugation, resuspended in resuspension buffer (25 mM Tris⋅HCl pH 8.0 and 300 mM NaCl) and lysed by sonication. Cell debris was removed by centrifugation and the cleared lysates were loaded onto TALON metal affinity resin (Clontech). The matrix was washed with 20 column volumes (CV) of resuspension buffer, followed by elution of bound proteins with elution buffer (25 mM Tris⋅HCl pH 8.0, 300 mM NaCl, and 500 mM imidazole). The eluted fractions were buffer-exchanged into resuspension buffer and incubated with Ulp1 protease at 4 °C overnight. Cleaved His₆-SUMO tags and uncut protein was removed via TALON metal affinity chromatography, and the cleaved target proteins were recovered in the flow-through. Purified protein was subjected to size exclusion chromatography on a HiLoad 16/600 Superdex 75 column (GE Healthcare) in gel-filtration buffer (25 mM Tris⋅HCl pH 8.0 and 300 mM NaCl). Purity of the proteins was analyzed by SDS/PAGE.

Circular Dichroism Spectroscopy.

Circular dichroism measurements were performed using a J-720 (JASCO) spectroscope. Initial spectra of Pa_FleQ^Full and mutants (5 μM in 20 mM phosphate buffer, pH 7.2; 0.1-cm light-path quartz cuvette) were obtained by averaging 16 scans over a wavelength range of 200–260 nm. To eliminate background effects, buffer contributions were subtracted in all measurements. Spectra were transformed to molar ellipticity (θ). Thermal denaturation analyses were carried out by measuring the ellipticity at 220 nm over a temperature range from 10 °C to 95 °C, at a constant heating rate of 10 °C/h. The fraction of denatured protein was calculated from the relationship (θ_n − θ_obs)/(θ_n − θ_d), where θ_obs is the ellipticity of the sample at a particular temperature and θ_d and θ_n are the values of ellipticity characteristic of the denatured and native state, respectively.

Crystallization, Data Collection, and Structure Determination.

Initial crystallization screens were conducted via sitting-drop vapor diffusion using commercial crystallization screens. Pa_FleQ^AAA crystallized at 18 °C against a solution of 0.1 M Bis-Tris pH 6.5, 0.2 M ammonium sulfate, and 25% (wt/vol) PEG3350. For crystallization of ADP- and ATP-γ-S–bound Pa_FleQ^AAA, the protein was preincubated with a fivefold molar excess of nucleotide for 2 h at 10 °C. Crystals of c-di-GMP–bound Pa_FleQ^AAA-HTH were obtained at 4 °C in the presence of 1 mM c-di-GMP in the drop and 0.05 M Bis-Tris pH 6.5, 30% (vol/vol) pentaerythritol ethoxylate (15/4 EO/OH), and 0.05 M ammonium sulfate in the mother liquor.

For data collection, crystals were soaked in mother liquor supplemented with cryo-protectants [up to 25% (wt/vol) xylitol or ethylene glycol] and flash-frozen in liquid nitrogen. Diffraction data were collected using either a Cu-rotating anode X-ray source (Pa_FleQ^AAA–ATP-γ-S) or synchrotron beamlines (beamline MX2 of the Laboratório Nacional de Luz Síncrotron, Campinas, Brazil, for apo-Pa_FleQ^AAA and Pa_FleQ^AAA-ADP; beamline A1 at Cornell High Energy Synchrotron Source, Cornell University, Ithaca, NY or beamline 24-ID at Advanced Photon Source, Argonne National Laboratories, Argonne, IL, for Pa_FleQ^AAA-HTH-c-di-GMP). All diffraction images were processed using the software package XDS (47). The apo-FleQ^AAA structure was solved using SAD with iodine as the anomalous scatterer, collected on crystals that were soaked in mother liquor supplemented with 500 mM NaI and 10% (vol/vol) ethylene glycol (48, 49). Initial phases were extended to 1.8 Å resolution from a native dataset using the PHENIX program package (50). ADP- and ATP-γ-S–bound Pa_FleQ^AAA were solved by molecular replacement in PHENIX, with the apo structure as the starting model. The structure of Pa_FleQ^AAA-HTH–c-di-GMP was solved by SAD phasing with data collected on crystals grown from selenomethionine-derivatized protein. The heavy atom positions were determined with SHELX (51) and automatic building was conducted in PHENIX. For all constructs, iterative model building and refinement in COOT (52) and PHENIX yielded the final structural models.

Electron Microscopy.

Pa_FleQ^Full (50 μM) was incubated with 100 μM ADP, 200 μM sodium vanadate, and 500 μM MgCl₂. Following a 30-min incubation on ice, the protein was gel-filtered against an ADP-vanadate–containing buffer [20 mM Tris⋅HCl pH 8.0, 100 mM NaCl, 2.5% (wt/vol) xylitol, 100 μM ADP, 200 μM sodium vanadate, and 500 μM MgCl₂] on a Superdex 200 10/300 GL size exclusion column (GE Healthcare). Five-microliter drops of protein at concentrations between 0.005 and 0.1 mg/mL were placed directly on glow-discharged carbon-coated grids (EMS) for 1 min. The grids were then blot-dried on filter paper and negatively stained with 2% (wt/vol) uranyl acetate in water. Specimens were examined on an FEI Tecnai T12 BioTWIN LaB6 electron microscope operating at 120 kV at nominal magnifications of 49,000 and 1–3 μm defocus. Images were recorded on a Gatan Ultrascan 4000 CCD camera. The contrast transfer function parameters were assessed using CTFFIND3 (53) and the phase flipping was done in SPIDER (54). Single particles were boxed manually and a single-particle stack was generated in EMAN2 (4,856 particles, 170 × 170-pixel box, 2.2 Å per pixel) (55). Normalization, centering, multireference alignment, multistatistical analysis, classification (∼40 particles per class), and rotational autocorrelation function calculation were done in IMAGIC-4D (Image Science Software).

Isothermal Titration Calorimetry.

Apparent dissociation constants (K_d) and stoichiometry of interaction (n) were measured by ITC using a VP-ITC calorimeter (Microcal). Calorimetric titrations of c-di-GMP (700 μM in the syringe; 7-μL injections) and wild-type or mutant Pa_FleQ^Full (30 μM in the cuvette) were carried out at 25 °C in assay buffer (25 mM Tris⋅HCl pH 8.0, 300 mM NaCl, 5 mM MgCl₂) with 150-s spacing between injections. ITC data were analyzed by integrating heat effects normalized to the amount of injected protein and curve-fitting based on a single-site binding model using the MicroCal Origin 7.0 software. Heat changes during ITC titrations likely combine contributions from c-di-GMP binding and accompanied changes in intra- and intermolecular protein interfaces. Heat from the dilution of c-di-GMP alone was subtracted before data analysis and the first injection was omitted from data processing.

In-Solution Chemical Cross-Linking.

For oligomeric state detection, protein in cross-linking buffer (25 mM Hepes pH 8.0 and 300 mM NaCl) was incubated with EGS at room temperature. Protein and cross-linker concentrations and incubation times are specified in the respective figure legends. Cross-linking reactions were quenched by adding 50 mM Tris⋅HCl pH 8.0, followed by incubation at 97 °C for 10 min in standard SDS/PAGE sample buffer. Oligomer formation was analyzed by SDS/PAGE.

ATPase Activity Assay.

ATPase activity of wild-type and mutant Pa_FleQ^Full was measured with the EnzChek Phosphate Assay kit (Life Technologies) (56). Assays were performed in triplicates in a 96-well plate according to the manufacturer’s instructions. Briefly, ATPase reactions (200 µL; 1 µM enzyme) were carried in ATPase buffer (25 mM Hepes pH 8.0, 300 mM NaCl buffer, 1 mM DTT, 10 mM KCl, 1 mM MgCl₂, and 1 mM ATP). The reactions were monitored at 1-min intervals at 360 nm in a microplate reader.

P. aeruginosa Strain Construction, RNA Isolation, and Real-Time Quantitative PCR.

P. aeruginosa strains PAO1ΔfleQΔfleN or strains containing different mutations in the fleQ gene were routinely grown in LB medium at 37 °C. Mutations in the fleQ gene were generated by an overlapping PCR procedure. Briefly, PCR fragments containing the mutated fleQ gene and flanking genomic DNA were cloned into pEX19Gm or miniCTX2. fleQ_WT and fleQ_R334E were cloned into miniCTX2. fleQ_I374E, fleQ_V256E, fleQ_R185E, fleQ_V380E, and fleQ_T149E were cloned into pEX19Gm. The resulting plasmids were then used to transform E. coli S17-1 strain and mobilized into the PAO1ΔfleQΔfleN strain. MiniCTX2 constructions allow the introduction of fleQ_WT and fleQ_R334E at the attB site on the P. aeruginosa chromosome, whereas pEX19Gm constructions allow the introduction of fleQ variant alleles at the fleQ native location. Plasmids pJN1120 (expressing YfiN/TpbB/PA1120 diguanylate cyclase) and pJN105 (empty vector) were electroporated into the constructed strains (14, 57, 58). Cell growth, RNA extraction, and RT-PCR were performed essentially as previously described (13, 14, 59). The pelA-forward (CCT TCA GCC ATC CGT TCT TCT) and pelA-reverse (TCG CGT ACG AAG TCG ACC TT) primers were used. Genomic DNA was used as a standard and transcript levels of all genes were normalized to total cDNA. Data presented are at least the average of two biological replicates. Error bars represent the SD between samples.