Diguanylate Cyclases Control Magnesium-Dependent Motility of Vibrio fischeri

Abstract

Flagellar biogenesis and hence motility of Vibrio fischeri depends upon the presence of magnesium. In the absence of magnesium, cells contain few or no flagella and are poorly motile or nonmotile. To dissect the mechanism by which this regulation occurs, we screened transposon insertion mutants for those that could migrate through soft agar medium lacking added magnesium. We identified mutants with insertions in two distinct genes, VF0989 and VFA0959, which we termed mifA and mifB, respectively, for magnesium-dependent induction of flagellation. Each gene encodes a predicted membrane-associated protein with diguanylate cyclase activity. Consistent with that activity, introduction into V. fischeri of medium-copy plasmids carrying these genes inhibited motility. Furthermore, multicopy expression of mifA induced other phenotypes known to be correlated with diguanylate cyclase activity, including cellulose biosynthesis and biofilm formation. To directly test their function, we introduced the wild-type genes on high-copy plasmids into Escherichia coli. We assayed for the production of cyclic di-GMP using two-dimensional thin-layer chromatography and found that strains carrying these plasmids produced a small but reproducible spot that migrated with an R_f value consistent with cyclic di-GMP that was not produced by strains carrying the vector control. Disruptions of mifA or mifB increased flagellin levels, while multicopy expression decreased them. Semiquantitative reverse transcription-PCR experiments revealed no significant difference in the amount of flagellin transcripts produced in either the presence or absence of Mg²⁺ by either vector control or mifA-overexpressing cells, indicating that the impact of magnesium and cyclic-di-GMP primarily acts following transcription. Finally, we present a model for the roles of magnesium and cyclic di-GMP in the control of motility of V. fischeri.

The limiting step in understanding signal transduction most often is the identification of the environmental signal that induces a physiological change. This has been true for two-component signaling (13, 58) and may also be true for the pathways that control the production of cyclic di-GMP (c-di-GMP) (45). This newly appreciated second messenger is synthesized from two GTP molecules by diguanylate cyclases (DGCs). These enzymes, found in numerous and diverse bacterial genomes, are readily identifiable through their signature GGDEF domains (19, 39, 40, 50, 53, 56). Furthermore, many bacterial species possess multiple proteins with domains that contain this GGDEF domain (for a recent review, see reference 45). Degradation of c-di-GMP is accomplished by phosphodiesterases containing either EAL or HD-GYP domains (8, 15, 49, 52, 57). Together, these activities maintain the steady-state concentration of c-di-GMP (46).

c-di-GMP first was discovered as a component of the cellulose biosynthesis enzyme complex from Gluconacetobacter xylinus, where it plays a vital role in promoting cellulose biosynthesis (46). c-di-GMP is now known to control exopolysaccharide production, rugose colony morphology, biofilm formation, and motility in a number of organisms (23, 27, 44, 53). In general, c-di-GMP production appears to promote a sessile, nonmotile lifestyle (45).

Numerous studies have shown that multicopy expression of DGCs inhibits motility (e.g., see references 3, 7, 34, and 53). However, the mechanism(s) by which this inhibition occurs has not been determined. In contrast, there exist a more limited number of studies in which mutations in DGC genes have been demonstrated to impact motility (1, 12, 23, 25, 43). Perhaps the best-characterized motility-associated DGC is the Caulobacter crescentus protein PleD, a two-component response regulator that contains a C-terminal GGDEF domain. During development, C. crescentus must eject its polar flagellum and grow a stalk in its stead. Flagellar ejection depends on cleavage of FliF, the protein that forms the MS ring at the base of the flagellum (1, 2, 23, 39). The ejection of the flagellum depends upon the production of c-di-GMP by PleD (1, 2, 23, 39).

Vibrio fischeri, naturally found free-living in seawater or in symbiotic association with the Hawaiian squid Euprymna scolopes (37), controls its flagellar biogenesis in response to magnesium (Mg²⁺) in the environment: when Mg²⁺ is abundant, as it is in seawater, V. fischeri cells possess flagella; when this cation is limiting, they do not (38). In contrast to PleD, loss of flagella in V. fischeri does not occur through ejection (38). To understand how Mg²⁺ controls flagellation, we sought mutants that could migrate through soft agar medium lacking added Mg²⁺. We report here the identification and characterization of two DGC genes, mifA and mifB, that inhibit migration in the absence of magnesium. We propose that these two DGCs constitute the keystone of an Mg²⁺-sensitive signaling pathway that regulates flagellar biogenesis in V. fischeri.

MATERIALS AND METHODS

Strains and media.

The strains used in this study are shown in Table 1. V. fischeri strain ESR1 (21) was used as the wild-type control for motility experiments with transposon insertion mutants. The wild-type V. fischeri strain ES114 (11) was used as the genetic background for construction of non-Tn-based mif mutants. Strain KV1421, a derivative of ES114 that contains a chromosomal erythromycin-resistant cassette (Tn7::erm) (32), was used as the wild-type control for motility experiments with erythromycin-resistant strains. E. coli strains Top10F′ (Invitrogen, Carlsbad, CA), CC118 λ pir (24), or TAM1 λ pir (Active Motif, Carlsbad, CA) were used for cloning. Triparental matings to introduce plasmid DNA into V. fischeri utilized Escherichia coli strain CC118λpir carrying the conjugation helper plasmid, pEVS104 (55).

TABLE 1.

Strains and plasmids used in this study

Strain or plasmid	Genotype	Source or reference
Strains
V. fischeri
ES114	Wild type	11
ESR1	Rifr	21
KV1421	Tn7::erm	This study
KV1897	RifrmifA::Tn10lacZ + delivery vector	This study
KV1902	RifrmifA::Tn10lacZ	This study
KV1903	RifrmifB::Tn10lacZ + delivery vector	This study
KV2530	mifC::pTMO123	This study
KV2532	mifB::pTMO125	This study
KV2672	mifA::pKV217	This study
KV2825	ΔmifB	This study
KV2826	ΔmifB mifA::pKV217	This study
E. coli
AJW678	thi-1 thr-1(Am) leuB6 metF159(Am) rpsL136 ΔlacX74	29
DH5α	endA1 hsdR17 (rK− mK+) supE44 thi-1 recA1 relA Δ(lacIZYA-argF)U169 phoA [φ80dlac Δ(lacZ)M15]	62
CC118 λ pir	Δ(ara-leu) araD Δlac74 galE galK phoA20 thi-1 rpsE rpsB argE(am) recA λ pir	24
TAM1 λ pir	mcrA Δ(mrr-hsdRMS-mcrBC) F80 lacZΔM15 ΔlacX74 recA1 araD139 Δ(ara-leu)7697 galU galK rpsL endA1 nupG λ pir	Active Motif, Carlsbad, CA
Plasmids
pAHMS16	pUC18-hmsT	28
pCR2.1 TOPO	Cloning vector, Kanr Apr	Invitrogen
pKV69	Low-copy vector, Cmr Tetr	60
pTMO105	mifA in pKV69 (pKV69 BamHI/SphI + 2.2-kb BamHI/SphI [mifA+] fragment from pTMO92; Tetr Cmr)	This study
pTMO126	mifA in pCR2.1 TOPO (pCR2.1 TOPO + 2,163-bp PCR fragment of mifA amplified with primers Up989Fwd and Down989Rev; Kanr Apr)	This study
pTMO142	mifCB in pCR2.1 TOPO (pCR2.1 TOPO + mifCB amplified with primers Upfus959F and Down959Rev; Kanr Apr)	This study
pTMO149	mifCB in pKV69 (pKV69 SmaI + BamHI/XbaI-filled fragment of mifCB from pTMO142; Tetr Cmr)	This study
pTMO155	mifB in pKV69 (pTMO149 with SalI deleted, which removes upstream sequences between the multiple cloning site and the 3′ end of mifC)	This study
pUC19	Cloning vector, Apr	63

V. fischeri strains were grown in the following media. For routine culturing, LBS medium (54) was used. For experiments, cells were grown in Tris-buffered saline (TBS), which contains 342 mM NaCl and 1% Bacto-tryptone (Becton Dickinson and Company, Sparks, MD) (17) or TBS-Mg²⁺ (TBS with 35 mM MgSO₄) (38). In some experiments, MgSO₄ was added to final concentrations between 35 and 200 mM, as indicated in the text. HEPES minimal medium (HMM) (47) containing 100 mM HEPES, pH 7.5, 0.3% Casamino Acids, and 0.2% glucose, was used for biofilm experiments. E. coli strains were grown in LB (16), brain heart infusion medium (Becton, Dickinson and Company, Sparks, MD), or SOC medium (51). For c-di-GMP assays in E. coli, MOPS (morpholinepropanesulfonic acid) medium was prepared as described by Neidhardt et al. (36), with the following exceptions: 0.4% glucose, 0.2 mM K₂HPO₄, and 1/50 the micronutrients were used. Where appropriate, antibiotics were added to the following final concentrations: rifampin at 100 μg/ml; tetracycline at 5 μg/ml; chloramphenicol at 5 μg/ml for V. fischeri and 25 μg/ml for E. coli; and erythromycin at 5 μg/ml for V. fischeri and 150 μg/ml for E. coli. Agar was added to a final concentration of 1.5% for solid media and either 0.225% or 0.25% for motility plates.

Cloning, sequencing, and mutant construction.

Plasmids constructed or used in this study are shown in Table 1 (and see data at http://www.meddean.luc.edu/lumen/DeptWebs/microbio/pub/pub.htm). Standard molecular biology techniques were used for all plasmid constructions. Restriction and modifying enzymes were purchased from New England Biolabs (Beverly, Mass.) or Promega (Madison, Wis.). DNA oligonucleotides used for amplifying mif genes (see supplemental material at http://www.meddean.luc.edu/lumen/DeptWebs/microbio/pub/pub.htm) were obtained from MWG Biotech (High Point, NC). The site of insertion of the transposon (Tn) in each mutant strain (KV1897, KV1902, and KV1903) was determined by cloning the Tn and flanking DNA to generate plasmids pTMO73, pTMO78, and pTMO85 (see data at http://www.meddean.luc.edu/lumen/DeptWebs/microbio/pub/pub.htm), respectively; plasmid pTMO85 was further subcloned to generate pTMO93. We sequenced DNA adjacent to the Tn in plasmids pTMO73, pTMO78, and pTMO93 with Tn- or plasmid-specific primers, using the services of Davis Sequencing, Inc. (Davis, CA).

We constructed vector-integration (Campbell insertion) (14) mutants by cloning an internal fragment of the appropriate genes into a suicide vector, pESY20, that cannot replicate in V. fischeri. The appropriate plasmid DNA was introduced into ES114 by triparental conjugation, followed by selection on erythromycin-supplemented LBS. The correct mutants were verified by Southern analysis. To construct the mifB deletion derivative, plasmid pKV231 (see data at http://www.meddean.luc.edu/lumen/DeptWebs/microbio/pub/pub.htm) was introduced into ES114. Cells containing the plasmid were selected on LBS containing chloramphenicol. The resulting colonies were passaged with or without chloramphenicol, and screened for increased motility on TBS soft agar plates. The presence of the mifB deletion was determined by PCR with primers complementary to regions flanking the insertion. The correctness of the resulting strain was verified by Southern analysis with a probe for the mifCB region. To construct the double mutant, the mifA Campbell construct, pKV217, was introduced into the mifB deletion strain by conjugation, followed by selection on LBS containing erythromycin. KV1421 was constructed by introducing plasmid pEVS107 (Tn7::erm) (32) into ES114 by conjugation in a mating that also included E. coli strains carrying the Tn7 transposase plasmid pUX-BF13 (6) and the conjugal plasmid pEVS104 (55).

Motility assays.

To assay motility of V. fischeri, individual mutants and control strains were grown to mid-exponential phase (optical density [OD] of ∼0.3), and 10-μl aliquots were spotted onto TBS or TBS-Mg²⁺ motility plates (containing as necessary the appropriate antibiotics). The diameter of migrating rings was measured over the course of 4 to 5 h of incubation at 28°C (17, 61). A similar strategy was used to assay motility of E. coli, except that the motility plates were based upon TB, which contains 86 mM NaCl and 1% Bacto-tryptone, and incubated for up to 12 h at 33°C.

Diguanylate cyclase activity assays.

E. coli cells were grown at 37°C in phosphate-limited MOPS medium supplemented with 45 μCi/ml of ³²P_i (obtained as carrier-free orthophosphate in dilute HCl, pH 2 to 3, from Amersham). One-hundred-microliter samples were collected in 1.5-ml Eppendorf Safe-Lock tubes (Fisher Scientific) to which 10 μl of cold (4°C) 11 N formic acid (88% formic acid; Fisher Scientific) had been aliquoted and incubated in an ice bath for 30 min. The unincorporated orthophosphate was precipitated with 16.5 μl Na-tungstate-tetraethylammonium chloride-procaine precipitate solution as described previously (10) and centrifuged at 14,000 × g for 15 min at 4°C. A 10-μl aliquot of the supernatant was immediately neutralized with 10 μl of 2-picoline, as described previously (9). The neutralized sample was processed by two-dimensional thin-layer chromatography (2D-TLC) on polyethyleneimine cellulose F plates (EMD Chemicals), as described previously (59). To avoid inconsistencies associated with difference of exposure, all plates from any given experiment were exposed simultaneously to the same PhosphorImager screen.

c-di-GMP-dependent phenotype analyses.

To assay the phenotypic consequences of c-di-GMP overproduction, we grew V. fischeri cells on chloramphenicol-containing LBS plates on which was spread 125 μl of a 0.2% stock of calcofluor (Fluorescent Brightener 28; Sigma Chemical, St. Louis, MO) dissolved in 1 M Tris, pH 9. We visualized fluorescence of the resulting colonies with UV light. We also examined cellulose production as described previously (22) using plates containing Congo red and Coomassie blue at final concentrations of 40 μg/ml and 15 μg/ml, respectively. To examine biofilm formation, cells were grown in HMM for 24 h with shaking. After staining with a 1% solution of crystal violet, the test tubes were rinsed three times with water and then photographed.

Western analysis.

Western blot analysis was performed as previously described (38) with rabbit anti-Vibrio parahaemolyticus flagellin antibody (33).

RT-PCR.

To analyze flagellin mRNAs from V. fischeri cells grown to early exponential phase (OD = 0.3 at 600 nm) in TBS or TBS-Mg²⁺, RNA was first extracted using the RNeasy Mini kit from QIAGEN (Valencia, CA). DNA contamination was removed by treatment with 5 U of RQ1 RNase-free DNase (Promega, Madison, WI) in 1× RQ1 DNase buffer for 2 h at 37°C, followed by phenol-chloroform extraction and ethanol precipitation. To make cDNA, 0.7 μg RNA was incubated with 11.5 μM random hexamer primers from IDT (Skokie, IL) at 75°C for 3 min and then cooled to 4°C, to allow primers to anneal to the RNA. Next, 1× Stratascript buffer, 0.5 μM deoxynucleoside triphosphates (dNTPs), and 50 U of Stratascript reverse transcriptase (Stratagene, La Jolla, CA) were added and cDNA was generated by incubation at 42°C for 1 h, followed by 5 min at 95°C. DNase contamination was determined by performing a mock cDNA reaction lacking reverse transcriptase and dNTPs. Then, gene-specific primers (see data at http://www.meddean.luc.edu/lumen/DeptWebs/microbio/pub/pub.htm) were used in a PCR (30 cycles of 94°C for 20 s, 54°C for 30 s, and 72°C for 1 min, followed by a final 5-min 72°C extension.) For flaA, flaB, flaD, and flaF, 2.5 μl cDNA was used as a template; for flaC, flaE, and S16 reactions, 2.5 μl 5× dilute cDNA was used. PCR mixtures contained 0.4 μM primers, 0.25 μM dNTPS, 1.5 mM MgCl₂, 1× Taq buffer, and 1 U Taq (Promega, Madison, WI). Reactions were run on 1.5% agarose-Tris-borate-EDTA gels and photographed using a charge-coupled device camera and AlphaEaseFC software (AlphaInnotech Corp., San Leandro, CA).

RESULTS

Negative regulation controls motility of V. fischeri.

Mg²⁺ controls migration of V. fischeri by promoting flagellar biogenesis (38). To account for this Mg²⁺-dependent migration, we hypothesized that Mg²⁺ could enhance the activity of a positive regulator or prevent the function of a negative regulator. The latter hypothesis predicts that it should be possible to identify mutants defective for a negative regulator of flagellar biogenesis. Such a mutant would exhibit increased migration in the absence of Mg²⁺. To test this possibility, we screened transposon mutant libraries for strains that exhibited enhanced migration on TBS (a tryptone-based medium lacking added MgSO₄) soft agar. In a screen of approximately 4,500 mutants, we identified 11 strains with increased migration on TBS. Careful measurements of migration rates of three of these revealed that they migrated substantially faster on TBS than ESR1, their parent (Fig. 1A and B and Table 1). In the absence of added Mg²⁺, displacement of the outer edge of the swarm produced by each of the mutants began within an hour of inoculation (data not shown). Furthermore, this outer edge quickly developed into a tight, migrating ring of cells that responded to thymidine (17) (Fig. 1A and data not shown). In contrast, wild-type displacement was delayed by at least 4 h and the characteristic ring of thymidine-responsive cells never fully formed. In the presence of Mg²⁺, both the migration and the thymidine-responsive rings of the three mutants were indistinguishable from those of their parent (Fig. 1C and D). These data are not consistent with the behavior of hypermotile mutants, but rather with a specific increase in migration caused by the disruption of a negative regulator. Since none of the mutants migrated to the same extent in the absence of Mg²⁺ as they did with Mg²⁺ addition (Fig. 1), we conclude that more than one gene product contributes to the inhibition of motility observed in the absence of Mg²⁺.

FIG. 1.

Motility of transposon mutants in the absence and presence of Mg²⁺. Exponential-phase cells of parent strain ESR1 (1; closed circles) and Tn insertion mutants KV1897 (2; open squares), KV1902 (3; closed squares), and KV1903 (4; open triangles), grown in TBS, were inoculated onto TBS or TBS-Mg²⁺ (containing 35 mM MgSO₄) soft agar plates (labeled −Mg²⁺ and +Mg²⁺, respectively) and incubated at 28°C. (A and C) Photographs of the migration of indicated strains after about 4 h of migration. (B and D) Migration was determined by measuring at hourly intervals the diameter of the outer migrating rings. At the last time point for the −Mg²⁺ condition, only a small percentage of ESR1 cells contributed to the ring formed; the majority remained in the spot at the center of the plate. The error bars represent the standard deviations of a representative experiment performed in triplicate. The same data for ESR1 are plotted on multiple panels for comparison. Note the difference in the y-axis scales of the graphs.

Two putative diguanylate cyclase genes control motility.

To understand the pathway of Mg²⁺-dependent migration control, we cloned and sequenced DNA flanking the sites of the three transposon insertions. Two of the strains (KV1897 and KV1902) contained insertions in VF0989 (Fig. 2A). The transposons mapped to the same location within VF0989, although one of the insertions also included sequences from the delivery vector. The third strain contained an insertion in VFA0959 (KV1903; Fig. 2B). In a subsequent Tn5 mutagenesis and motility screen to identify additional genes involved in Mg²⁺-dependent migration control, we isolated 17 more mifA mutants, at least 11 of which were independently derived. These data further support the identification of mifA as an important and specific regulator of Mg²⁺-dependent migration.

FIG. 2.

Schematic representation of the mifA, mifCB, and flanking genes. (A) The mifA (VF0989) gene and genes flanking it on V. fischeri chromosome 1 are depicted as arrows. (B) The mifC (VFA0960) and mifB (VFA0959) genes and the genes flanking them on V. fischeri chromosome 2 are depicted as arrows. The numbers below indicate the spacing between the genes.

Both VF0989 and VFA0959 encode proteins with C-terminal GGDEF domains (GGEEF in the case of VF0989), indicating that they may encode DGCs. The amino termini of the predicted proteins each contain two putative membrane-spanning domains and a potential periplasmic loop. Finally, VFA0959 contains a predicted PAS sequence. Immediately upstream of VFA0959 is an open reading frame (VFA0960) that overlaps by 1 bp; thus, VFA0959 and VFA0960 likely constitute an operon (Fig. 2B). VFA0960 encodes a predicted periplasmic protein. Due to the roles of these genes in Mg²⁺-dependent induction of flagellar biogenesis (along with supporting documentation described below), we designate VF0989 as mifA, VFA0959 as mifB, and VFA0960 as mifC.

Control of motility involves MifA and MifB.

To begin to elucidate the roles of mifA, mifB, and mifC in Mg²⁺-dependent migration and to rule out any potential complications from the study of transposon insertion mutants (derived from the rifampin-resistant strain ESR1), we used a vector-integration approach (18) to disrupt each of these genes in the wild-type strain ES114. Because we had only obtained the mifB mutant once, we particularly sought to verify its role in Mg²⁺-dependent migration. In the course of these studies, we determined that the erm gene (in the vector portion of the insertion) in V. fischeri caused a slight motility defect; therefore, where appropriate for subsequent experiments, we used as our control KV1421, a derivative of ES114 that contains the erm gene inserted at the Tn7 site. KV1421 and ES114 exhibited two behaviors that distinguished them from ESR1; both were observable in the presence of Mg²⁺. First, a substantial proportion of KV1421 (and ES114) cells remained at the site of inoculation (compare Fig. 3C to Fig. 1C; data not shown). Second, the characteristic inner band of cells that migrate in response to serine (17) appeared more rapidly with KV1421 and ES114 (Fig. 3C and data not shown) than it did with ESR1 (data not shown). As predicted from our original results (Fig. 1), however, disruption of either mifA or mifB allowed the cells to migrate in the absence of Mg²⁺ but did not substantially impact migration in its presence (Fig. 3B and D and data not shown). These data confirmed specific roles for mifA and mifB in the inhibition of migration in the absence of Mg²⁺.

FIG. 3.

Effects on motility of mutations in mif genes. Migration of mif mutants was assayed as described in the legend to Fig. 1. Photographs of the migration of indicated strains were taken after about 5 h of migration. Strains include the Erm^r control KV1421 (1; closed circles), KV2532 (mifB) (2 and 4; open triangles), KV2530 (mifC) (3; open diamonds), KV2826 (mifA mifB) (5; open circles), and KV2672 (mifA) (6; open squares). The error bars represent the standard deviations of a representative experiment performed in triplicate. In the absence of Mg²⁺, only a small percentage of KV1421 cells contributed to the ring formed at the last time point; the majority remained in the spot at the center of the plate. The same data for KV1421 are plotted on multiple graphs for comparison.

Due to the putative operon structure comprising mifC and mifB, we also asked whether mifC contributed to Mg²⁺-dependent induction of motility. Indeed, a strain in which mifC was insertionally inactivated through vector integration displayed motility characteristics typical of the mifB mutant (Fig. 3). These data suggest that mifC is important, at least for providing a promoter for mifB transcription and potentially as a component in its own right. A more precise characterization of the role of mifC will depend upon the construction of unmarked nonpolar mutations. Finally, we constructed a double mutant containing an insertion in mifA and an unmarked deletion in mifB. Relative to the parent strains, the double mutant exhibited a slight but consistent increase in migration in the absence of Mg²⁺ (Fig. 3A and B and data not shown). Because the loss of both MifA and MifB does not bring migration to the same level as does the addition of Mg²⁺, we conclude that another Mif component(s) likely exists.

Multicopy expression of mifA or mifB inhibits motility and promotes biofilm formation.

To further understand the functions of MifA, MifB, and MifC, we cloned wild-type copies of mifA and the putative mifCB operon and introduced them on a medium copy plasmid (about 10 copies per cell; E. Stabb, personal communication) into wild-type or transposon mutant strains. This multicopy expression reduced the migration of the resulting cells, relative to the vector control, regardless of the strain background (e.g., wild-type and mifA or mifB mutant strains; Fig. 4 and data not shown). Microscopic examination showed that multicopy expression of mifA and mifCB also reduced the percentage of motile cells relative to the vector control (data not shown). This result is similar to that observed for other bacteria in which DGCs are overexpressed (for examples, see references 3, 7, 34, and 53). On motility plates, the inhibition of migration caused by multicopy mifA expression was not substantially overcome by Mg²⁺ addition, as the vast majority of the cells remained at the site of the initial inoculation (Fig. 4) (data not shown). This was true even when the concentration of Mg²⁺ was increased above the optimal 35 mM level, to upwards of 200 mM (data not shown). To clarify the role of MifB, we removed most of the upstream mifC sequences from the mifCB plasmid construct and asked whether it retained the ability to inhibit motility. We found that multicopy expression of mifB alone similarly decreased migration of the wild-type strain in the absence of Mg²⁺. Intriguingly, however, addition of Mg²⁺ restored full motility to strains carrying multicopy mifB. These data suggest a role for MifC in modulating the activity of MifB.

FIG. 4.

Effects on motility of multicopy mifA, mifCB, and mifB. Migration of wild-type strain ES114 carrying the vector control (pKV69; closed circles) or plasmid pTMO105 (pmifA; open squares), pTMO149 (pmifCB; open triangles), or pTMO155 (pmifB; filled triangles) was assayed as described in the legend to Fig. 1, except that chloramphenicol was added at a final concentration of 2.5 μg/ml.

Because the observed motility phenotype of wild-type cells carrying either multicopy mifA or mifCB resembled that of other overexpressed DGCs, we asked whether these strains produced other phenotypes known to be correlated with DGC activity, including cellulose biosynthesis, biofilm formation, and wrinkled colony morphology. First, we tested cellulose biosynthesis by examining the ability of V. fischeri cells to fluoresce under UV light on plates containing the cellulose-binding dye, calcofluor. We found that multicopy expression of mifA induced bright fluorescence, but mifCB and the vector control did not (Fig. 5A). These data indicate that MifA induced an alteration in the sugar composition of the cell surface. We further examined this phenotype using Congo red plates, another indicator of cellulose biosynthesis, and similarly found that multicopy mifA increased the dye binding properties of the cell (Fig. 5B). Second, we tested biofilm formation by cells carrying mifA, mifCB, or the vector control by growing cells in a minimal medium, followed by staining adherent material with crystal violet. Consistent with the increased cellulose biosynthesis, multicopy expression of mifA, but not mifCB, substantially enhanced biofilm formation (Fig. 5C and data not shown). Finally, we investigated colony morphology of cells containing multicopy mifA, mifCB, or the vector control. After prolonged growth at room temperature, cells containing multicopy mifA formed wrinkled colonies, whereas cells carrying multicopy mifCB or the vector control produced smooth, round colonies characteristic of the wild-type strain (Fig. 5D and data not shown). Together, these data indicate that whereas multicopy expression of mifCB impacts only motility, multicopy mifA both inhibits motility and alters other cell surface characteristics.

FIG. 5.

Phenotype analysis of V. fischeri overexpressing mifA or mifCB. Cells of wild-type ES114 carrying vector (pKV69), multicopy mifA (pmifA; pTMO105), or multicopy mifCB (pmifCB; pTMO149) were examined for phenotypes consistent with increased c-di-GMP production. (A) Strains were grown on LBS plates with calcofluor and then exposed to UV light. (B) Strains were grown on LBS plates containing Congo red and Coomassie blue. (C) Strains were grown in HMM with shaking for 24 h and then stained with crystal violet as described in Materials and Methods. (D) Colony morphology of the vector control and multicopy mifA strains after prolonged growth on the LBS plates shown in panel B.

MifA and MifB exhibit diguanylate cyclase activity.

To further elucidate the roles of MifA and MifB, we assayed their activity in the heterologous microbe, E. coli. First, we introduced high-copy plasmids containing either mifA or mifCB and evaluated their effect on migration of E. coli on soft agar. Both mifA and mifCB (Fig. 6) decreased migration of E. coli, although not to the same extent as similar plasmids did to V. fischeri (Fig. 4). Then, we assayed potential DGC activity in these strains by separating small phosphorylated compounds by 2D-TLC. The mifA and mifCB constructs induced the production of a spot that migrated with an R_f value similar to that of the positive control, E. coli carrying the DGC HmsT from Yersinia pestis (Fig. 6B and C and data not shown). This spot was absent from the vector controls (Fig. 6D and data not shown). Quantification of these spots revealed that MifA produced more c-di-GMP than MifCB (Fig. 6E). Our data are consistent with the notion that MifA and MifCB harbor DGC activity (see Discussion).

FIG. 6.

Production of c-di-GMP by MifA and MifB. (A) Multicopy expression of mifA and mifCB inhibits migration of E. coli. Aliquots (5 μl) of E. coli strain AJW678 carrying pTMO126 (pmifA; open squares), pTMO142 (pmifCB; open triangles), or the vector control (pCR2.1 TOPO; closed circles), grown overnight in TB, were spotted onto the surface of TB motility plates and incubated at 34°C. The experiment was performed in triplicate at least twice. Error bars indicate standard deviation. (B to D) Multicopy expression of mifA in E. coli results in the accumulation of c-di-GMP. Cells of E. coli strain AJW678 carrying (B) pAhms16 (phmsT), (C) pTMO126 (pmifA), or (D) pUC19 (vector) were grown at 37°C in low-phosphate MOPS medium supplemented with 0.4% glucose, harvested during mid-exponential growth, and subjected to formic acid extraction. Small phosphorylated molecules were separated by 2D-TLC as described in Materials and Methods. The arrowheads indicate the signal corresponding to c-di-GMP or its position. The experiment was performed at least twice, each time with multiple samples. (E) Histogram summarizing c-di-GMP production as a percentage of phosphorylated compounds in extracts of E. coli cells from the data of panels B to D and data not shown. 1, vector (pCR2.1 TOPO); 2, pmifCB (pTMO142); 3, pmifA (pTMO126); 4, vector (pUC19); and 5, phmsT (pAhms16). The error bars represent standard deviation.

mifA and mifCB impact flagellin levels.

Previously, we reported that V. fischeri cells grown in the absence of Mg²⁺ had little flagellin protein and either were aflagellate or contained, at most, a single flagellum (38). Since the mif mutants were capable of migrating in the absence of Mg²⁺, we hypothesized that disruption of mif genes promotes migration by permitting the synthesis and assembly of functional flagella, which are composed primarily of flagellins. To determine whether mif pathway components contributed to the decreased flagellin levels in wild-type cells grown without Mg²⁺, we probed whole-cell extracts from cells grown with or without Mg²⁺ using an anti-V. parahaemolyticus flagellin antibody (33). This antibody has been previously shown to recognize at least five of the six distinct flagellin proteins (FlaA to -E) produced by V. fischeri (35, 38). Consistent with the observed motility phenotypes, when grown in the absence of Mg²⁺, strains with mutations in mifA, mifB, or both exhibited an increase in flagellin levels relative to the wild-type strain; when grown with Mg²⁺, they produced equivalent levels of flagellins (Fig. 7A). In contrast, multicopy expression of either mifA or mifCB substantially reduced the levels of flagellins produced by cells, in both the absence and presence of Mg²⁺ (Fig. 7B and C). These data are consistent with a model that has MifA and MifB influencing motility by controlling flagellin levels.

FIG. 7.

Immunoblot analysis of mif mutant and overexpression strains. Flagellin proteins extracted from cells grown to mid-exponential phase in TBS or TBS-Mg²⁺ were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, transferred to a membrane, and detected by a V. parahaemolyticus antiflagellin antibody. The strains used in panel A were ES114 (lanes 1 and 5), KV2532 (mifB; lanes 2 and 6), KV2672 (mifA; lanes 3 and 7), and KV2826 (mifA mifB; lanes 4 and 8). Panels B and C were derived from ES114 with (lanes 1) vector pKV69, (lanes 2) pTMO105 (pmifA), or (lanes 3) pTMO149 (pmifCB). Strains were grown in TBS with or without Mg²⁺, as indicated.

mifA and mifCB appear to act posttranscriptionally.

In all bacteria studied extensively, the flagellin genes reside at the bottom of the flagellar transcriptional hierarchy, potentially due to the need to tightly control these highly expressed genes (31, 42). However, many of the characterized pathways for c-di-GMP control appear to act posttranscriptionally (27, 41, 45). We therefore analyzed whether environmental Mg²⁺ or mif pathway components impacted steady-state levels of flagellin transcripts, using semiquantitative RT-PCR (Fig. 8). First, we found that the Mg²⁺ content of the medium did not substantially alter the levels of flagellin transcripts. Then, we determined that they were largely unaltered by the presence of multicopy mifA, despite the fact that this plasmid abolishes motility (Fig. 4) and dramatically reduces flagellin protein levels (Fig. 7B and C). Taken together, these data argue against a model for mif pathway control of flagellar biogenesis that relies primarily on a transcriptional mechanism.

FIG. 8.

Regulation of flagellin gene transcription by Mg²⁺ and multicopy expression of mifA. The gene amplified in each RT-PCR is labeled to the left of the panel. S16 is a ribosomal protein gene whose levels would be predicted to be unchanged by Mg²⁺ or MifA and was therefore used as a control for the efficiency of cDNA production. S16-RT represents reactions performed using the mock cDNA generated in the absence of reverse transcriptase and control for the presence of DNA contamination in the RNA. N, negative control (dH₂O as PCR template); P, positive control (genomic DNA as PCR template). − and +, presence and absence, respectively, of 35 mM MgSO₄ in the growth media.

DISCUSSION

We previously reported that flagellation and motility of V. fischeri depended upon Mg²⁺ in the medium. Here, we describe two novel flagellar regulators, MifA and MifB, that contribute to the poor motility of V. fischeri in the absence of Mg²⁺. When either mifA or mifB was disrupted, the cells exhibited a substantial increase in migration on TBS soft agar plates. Because wild-type cells under these conditions are largely aflagellate, the increase in migration of mif mutants most likely stems from an increase in the number of assembled flagella. In support of this conclusion, these mutations caused an increase in the levels of flagellin proteins in cells grown without Mg²⁺.

Both MifA and MifB are predicted to contain conserved GGDEF domain sequences, indicating that they likely function as DGCs. For MifA, our experiments demonstrate such a role, as follows: (i) E. coli strains carrying mifA on a multicopy plasmid synthesized a molecule that migrated on 2D-TLC plates with an R_f value similar to the positive control (Fig. 6); (ii) multicopy mifA expression inhibited motility of V. fischeri, and to a lesser extent, E. coli (Fig. 4 and 6A); and (iii) multicopy mifA expression promoted some of the same phenotypes induced by known DGCs, including increased cellulose biosynthesis and biofilm formation (Fig. 5). For MifB, our experiments are less conclusive: the impact of multicopy mifCB was primarily at the level of inhibiting motility; neither biofilm formation nor cellulose biosynthesis was noticeably impacted, and c-di-GMP production was difficult to discern. Like MifA, however, MifB contains a GGDEF domain and neither protein contains an EAL domain that would promote c-di-GMP degradation. Similar to that observed for mifA mutants, mifB mutants exhibited an increase in Mg²⁺-independent migration. Finally, like multicopy mifA, multicopy mifCB reduced flagellin levels of wild-type cells. Together, these data suggest that MifB functions like MifA, albeit with lower activity. This weaker effect could be due to differences in expression or activity of the protein expressed from the mifCB plasmid tested. Indeed, mifC and mifB are likely to be translationally coupled, and our overexpression data suggest that MifC influences the activity of MifB.

V. fischeri is predicted to contain about 30 genes with GGDEF domains (48). This abundance suggests a network of pathways that integrates multiple signals into a single second messenger (20). Alternatively, single pathways or subsets of pathways might work in relative isolation due to their localization or the existence of microenvironments (25, 39, 53). Our screens for Tn insertion mutants with increased motility in the absence of Mg²⁺ consistently yielded insertions in the mifA gene. These data suggest that MifA plays a specific role in flagellar biogenesis and not a more general role as a contributor to the intracellular pool of c-di-GMP. To date, we have identified only one insertion mutant disrupted for mifB; however, the phenotypes of the original and two newly constructed mifB mutants exhibit motility phenotypes similar to that of the mifA mutants, suggesting that MifB also is a specific component of the Mif pathway. MifA and MifB each contain a predicted periplasmic loop, which potentially could permit membrane localization. If these proteins were directed to the flagellated cell pole, the localized production of c-di-GMP could be an efficient means of specifically interfering with flagellar biogenesis and not other cellular processes. Such localization might, in fact, be critical: the relatively insensitive TLC assay failed to detect c-di-GMP production by V. fischeri cells carrying multicopy mifA or mifCB (A. Klein and A. Wolfe, unpublished data), despite the fact that these plasmids profoundly impacted motility. Where these proteins are localized and whether localization is critical to Mg²⁺-dependent motility will be the subject of further research.

The mechanisms used by c-di-GMP to influence behavior remain obscure, although the recent identification of a putative c-di-GMP binding domain, PilZ, has substantially advanced our understanding of this small molecule (4). Similar to cAMP, which requires only the transcription factor CRP (also known as CAP) (5, 30), c-di-GMP is predicted to work by direct interaction with its targets (19). Such is the case with the biosynthesis of cellulose, an extracellular polysaccharide (EPS), in G. xylinus and Salmonella enterica serovar Typhimurium (44). In these organisms, multiple DGCs and phosphodiesterases regulate the intracellular concentration of c-di-GMP, which binds directly to a cellulose synthesis complex that includes BcsA, a glycosyltransferase that possesses a predicted PilZ domain (4).

Like cellulose synthesis, several c-di-GMP-associated behaviors (e.g., ejection of the C. crescentus flagellum, the hemin storage [Hms] phenotype of Yersinia pestis, and twitching motility in Pseudomonas aeruginosa) appear to be regulated posttranslationally (1, 23, 25, 28, 41). Our work to date similarly suggests a mechanism of posttranscriptional control by Mg²⁺ and c-di-GMP: (i) Mg²⁺ does not substantially impact levels of any of the six flagellin transcripts (Fig. 8), but does impact flagellin levels (38) (Fig. 7A); (ii) in the absence of Mg²⁺, cells disrupted for either mifA or mifB exhibit increased motility (Fig. 1 and 3) and flagellin levels (Fig. 7A) but no changes in the levels of flagellin transcripts (Fig. 8); and (iii) in the absence of Mg²⁺, multicopy mifA nearly abolishes motility (Fig. 4) and substantially decreases flagellin levels (Fig. 7B), but does not significantly alter flagellin transcripts (Fig. 8). Our work has not yet fully elucidated the mechanism by which loss of motility occurs in the absence of Mg²⁺; however, because the levels of flagellin proteins are impacted, Mg²⁺ and/or c-di-GMP may affect translation or protein stability. Intriguingly, multicopy expression in C. crescentus of constitutively active versions of the P. aeruginosa DGC WspR inhibited motility but apparently not flagellation (3). The difference between this behavior and the one we describe in the present report indicates that the level at which large amounts of c-di-GMP impact motility may not be universal.

On the basis of the data presented here, we have constructed a model for known and predicted components of the mif pathway (Fig. 9). We propose that, in the absence of Mg²⁺, MifA and MifB use GTP to produce c-di-GMP, which inhibits flagellation, potentially through inhibiting translation, enhancing degradation of flagellar proteins, or inhibiting assembly. In the presence of Mg²⁺, flagellation may occur if MifA and MifB become inactivated and thus fail to produce c-di-GMP or, alternatively, if c-di-GMP degradation is increased. A periplasmic binding protein (e.g., MifC) may modulate the activities of these proteins.

FIG. 9.

Model for Mg²⁺-dependent induction of flagellation. In the absence of Mg²⁺, MifA and MifB use GTP to produce c-di-GMP, which inhibits flagellation, potentially through (A) inhibiting translation, (B) enhancing degradation of flagellar proteins, or (C) inhibiting assembly. In the presence of Mg²⁺, flagellation may occur if MifA and MifB become inactivated and thus fail to produce c-di-GMP, or, alternatively, if c-di-GMP degradation is increased. A periplasmic binding protein (e.g., MifC [gray oval]), may modulate the activities of these proteins.

It remains a formal possibility that Mg²⁺ and Mif form parallel pathways that each impact flagellation, such that the cation does not influence the activity of the Mif proteins. For example, Mg²⁺ may enhance flagellin synthesis through another mechanism, while Mif could potentially control degradation regardless of Mg²⁺ levels. In this scenario, the influence of the loss of Mif would be most discernible when little flagellin is being synthesized. Such a possible alternative could be supported by some of the data that we collected. First, mutation of both mifA and mifB together failed to restore the same level of motility that occurs in the presence of Mg²⁺. Second, multicopy expression of mifA inhibited migration of V. fischeri, even in the presence of 200 mM Mg²⁺. Alternatively, there may be missing components, such as an Mg²⁺-responsive component. Indeed, the activity of MifB and its responsiveness to Mg²⁺ vary, depending on the presence of MifC. In addition to a Mg²⁺-responsive component, we predict the involvement of a phosphodiesterase, as such proteins have been shown to play important roles in controlling c-di-GMP levels (8, 15, 57) and flagellar biogenesis (26); the activity of this protein could be impacted by Mg²⁺ levels. Finally, we also predict the existence of a c-di-GMP-binding protein, such as a PilZ domain protein (4), that interferes with translation or protein stability or prevents flagellar assembly. Intriguingly, V. fischeri contains genes that encode four PilZ domain proteins, one of which, VF1838, is situated between the flagellar export genes flhA and flhB. This gene is a prime candidate for playing a role in flagellar assembly. Our screens have yielded 10 additional mutants for which we have not yet determined the location of the Tn (T. O'Shea, A. Wolfe, and K. Visick, unpublished data); the identification of the disrupted genes may allow us to refine our model for the role of Mg²⁺ and Mif in promoting motility.

In summary, this work establishes a novel pathway that impacts Mg²⁺-sensitive motility through an apparent posttranscriptional mechanism. Several studies have shown that overexpression of c-di-GMP inhibits motility in other organisms. However, this study is among the first to identify an environmental signal and discrete regulators specifically involved in flagellar control.