Conserved Residues in the HAMP Domain Define a New Family of Proposed Bipartite Energy Taxis Receptors

Abstract

HAMP domains, found in many bacterial signal transduction proteins, generally transmit an intramolecular signal between an extracellular sensory domain and an intracellular signaling domain. Studies of HAMP domains in proteins where both the input and output signals occur intracellularly are limited to those of the Aer energy taxis receptor of Escherichia coli, which has both a HAMP domain and a sensory PAS domain. Campylobacter jejuni has an energy taxis system consisting of the domains of Aer divided between two proteins, CetA (HAMP domain containing) and CetB (PAS domain containing). In this study, we found that the CetA HAMP domain differs significantly from that of Aer in the predicted secondary structure. Using similarity searches, we identified 55 pairs of HAMP/PAS proteins encoded by adjacent genes in a diverse group of microorganisms. We propose that these HAMP/PAS pairs form a new family of bipartite energy taxis receptors. Within these proteins, we identified nine residues in the HAMP domain and proximal signaling domain that are highly conserved, at least three of which are required for CetA function. Additionally, we demonstrated that CetA contributes to the invasion of human epithelial cells by C. jejuni, while CetB does not. This finding supports the hypothesis that members of HAMP/PAS pairs possess the capacity to act independently of each other in cellular traits other than energy taxis.

HAMP domains (named for their presence in histidine kinases, adenylyl cyclases, methyl-accepting chemotaxis proteins, and phosphatases) (6) represent a common element in numerous bacterial signal transduction proteins. More than 11,700 known or predicted proteins containing HAMP domains are identified in the SMART database (39). The vast majority of these proteins are eubacterial, but HAMP domains have also been identified in archaea and lower eukaryotic organisms. HAMP domains are thought to play a role in intramolecular communication between the input and output domains of a single protein (4-6). HAMP domains have been studied predominantly in transmembrane receptors that translate a signal originating extracellularly to an intracellular signal transduction domain.

Our understanding of how HAMP domains function has been hampered by considerable sequence divergence among these domains and a paucity of structural data. Sequence analysis and mutagenesis studies have indicated that HAMP domains consist of two amphipathic helices (AS-1 and AS-2), which are joined by a flexible loop region to form a coiled-coil (5, 14, 47). Recently, the structure of the HAMP domain from the Archaeoglobus fulgidus protein Af1503 was solved (31). Af1503 is atypical of HAMP domain-containing proteins in that it lacks an output signal transduction domain (31). This structure consists of two amphipathic helices that come together in a parallel coiled-coil. These helices form a four-helix bundle in a HAMP domain dimer. This four-helix bundle adopts an unusual knobs-to-knobs conformation. These findings gave rise to a model where a shift in two transmembrane helices is translated into a gear-like 26° rotation of the helices relative to one another within the HAMP dimer four-helix bundle (31).

HAMP domains are often found in transmembrane receptors with extracellular input and intracellular output domains, although this is not true of all HAMP domains. A search of the SMART database reveals more than 400 known or predicted proteins with a HAMP domain that lack a predicted transmembrane domain. Even accounting for the fact that some of these proteins may actually have transmembrane domains that the SMART tool missed, there are clearly a number of proteins in nature that have HAMP domains and lack transmembrane domains. The mechanism by which HAMP domains might function in such proteins has not been extensively probed. Studies of a HAMP domain-containing protein in which the input and output signals both occur in the cytoplasm are limited to Escherichia coli Aer. Aer, the major energy taxis receptor of E. coli (10, 45), possesses four major domains: (i) a PAS domain (named after the three proteins [Per, ARNT, and Sim] where it was first identified) (51) that binds flavin adenine dinucleotide (FAD), the redox state of which is thought to reflect the redox state of an element(s) of the electron transport system, (ii) two transmembrane domains separated by a short periplasmically accessible region, (iii) a HAMP domain, and (iv) a conserved signaling domain present in all methyl-accepting chemotaxis proteins (MCPs) (1, 50). The PAS domain of Aer has been predicted to interact directly with the HAMP domain to transmit an energy taxis signal parallel to, rather than across, the inner membrane (50).

An energy taxis system consisting of a variation on the domain arrangement of Aer was previously identified in Campylobacter jejuni (27). C. jejuni, a microaerophilic, gram-negative bacterium commonly found in the gastrointestinal tracts of chickens and other livestock, is one of the most common causes of food-borne gastroenteritis in the United States. The flagellar motility of this bacterium has proven essential for both its commensal and its pathogenic lifestyle (25, 57). An energy taxis system of C. jejuni was identified in a screen of a transposon library for mutants defective in flagellar motility (27). This system consists of two proteins, CetA and CetB (formerly known as Cj1190c and Cj1189c, respectively), which together contain all of the domains of the single protein Aer. CetA, a predicted membrane-bound protein, possesses a predicted HAMP domain and the signaling domain. CetB, a predicted cytoplasmic protein, possesses a predicted PAS domain (27).

CetA and CetB are proposed to interact with one another directly to transduce an energy taxis signal via a mechanism similar to that of the single protein Aer (27). Since the HAMP domain of Aer is proposed to interact directly with the PAS domain, we hypothesize that the HAMP domain of CetA may mediate an interaction between CetA and CetB. Separation of these domains into distinct proteins may enable CetA and/or CetB to interact with other proteins and participate independently in alternate signaling pathways (27).

In this study, we determined that the HAMP domain of CetA differs from that of Aer in predicted secondary structure. Based on similarity with the CetA HAMP domain, we identified other members of a new family of putative bipartite energy taxis transducers. We found that the CetA homologs in this family possess highly conserved HAMP domain residues, at least three of which are required for wild-type function of CetA in energy taxis. Finally, we determined that the ΔcetA mutant, but not the ΔcetB mutant, has a defect in invasion of human epithelial cells, supporting the hypothesis that CetA and/or CetB may function independently of one another in cellular processes other than energy taxis.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

All bacterial strains and plasmids used in this study are listed in Table 1. DRH212, a spontaneous streptomycin-resistant mutant of the clinical isolate C. jejuni 81-176 (27), was the background strain for all mutants studied and is referred to as the wild type. C. jejuni was routinely grown on Mueller-Hinton (MH) agar with 10 μg/ml trimethoprim under microaerobic conditions (85% N₂, 10% CO₂, 5% O₂) in a tri-gas incubator. For C. jejuni, the following antibiotic concentrations were used: 10 μg/ml trimethoprim, 30 μg/ml cefoperazone, 50 μg/ml kanamycin, 20 μg/ml chloramphenicol, and 0.1 or 2 mg/ml streptomycin. E. coli was grown in Luria-Bertani (LB) agar or broth. For E. coli, the following antibiotic concentration was used: 50 μg/ml kanamycin or 100 μg/ml ampicillin.

TABLE 1.

Bacterial strains and plasmids

Strain or plasmid	Relevant characteristics	Source or reference
Bacteria
E. coli
JM101	F′ traD36 proA+B+ lacIq Δ(lacZ)M15/Δ(lac-proAB) glnV thi	New England Biolabs
DH5α/pRK212.1	Contains conjugative plasmid for transfer of plasmid into C. jejuni	20
C. jejuni
DRH212	81-176 Smr; spontaneous mutant	27
DRH304	cetB::cat-rpsL; intermediate strain for deletion mutagenesis	27
DRH307	ΔcetB	27
DRH321	ΔrpoN	27
DRH333	ΔcetA	27
KYCj172	ΔcetAB	This study
Plasmids
pUC19	Ampr	New England Biolabs
pRY108	Kmr; E. coli/C. jejuni shuttle vector	56
pKTY60	pUC19 with a 3.5-kb fragment containing cj1191c, cetA (cj1190c), and cetB (cj1189c) regions cloned into the KpnI site	This study
pKTY62	pKTY60 with a deletion from the first codon of cetA to the last codon of cetB	This study
pKTY152	pKTY60 with the D94A mutation in the cetA coding sequence	This study
pKTY153	pKTY60 with the E102A mutation in the cetA coding sequence	This study
pKTY154	pKTY60 with the E97A mutation in the cetA coding sequence	This study
pKTY155	pKTY60 with the K118A mutation in the cetA coding sequence	This study
pKTY156	pKTY60 with the R101A mutation in the cetA coding sequence	This study
pKTY157	pKTY60 with the R117A mutation in the cetA coding sequence	This study
pKTY158	pKTY60 with the R71A mutation in the cetA coding sequence	This study
pKTY159	pKTY60 with the Y116A mutation in the cetA coding sequence	This study
pKTY160	pKTY60 with the Y99A mutation in the cetA coding sequence	This study
pKTY360	pRY108 with a 2.4-kb fragment containing the cetA and cetB coding sequences cloned into the XmnI site	This study
pKTY361	pKTY360 with the R71A mutation in the cetA coding sequence	This study
pKTY362	pKTY360 with the D94A mutation in the cetA coding sequence	This study
pKTY363	pKTY360 with the E97A mutation in the cetA coding sequence	This study
pKTY364	pKTY360 with the Y99A mutation in the cetA coding sequence	This study
pKTY365	pKTY360 with the R101A mutation in the cetA coding sequence	This study
pKTY366	pKTY360 with the E102A mutation in the cetA coding sequence	This study
pKTY367	pKTY360 with the Y116A mutation in the cetA coding sequence	This study
pKTY368	pKTY360 with the R117A mutation in the cetA coding sequence	This study
pKTY369	pKTY360 with the K118A mutation in the cetA coding sequence	This study

Bioinformatic analysis.

The HAMP domain in CetA was identified by PSI-BLAST (2). Secondary-structure predictions were obtained using the PSA server (http://bmerc-www.bu.edu/psa/request.htm). Proteins that contained a HAMP domain and proximal signaling domain similar to those of CetA (residues 44 to 139) were identified by a BLAST search of the nonredundant protein database. Those proteins identified in this BLAST search that also possessed a neighboring open reading frame (ORF) containing a PAS domain were further analyzed for the presence of other functional domains using the SMART tool (39). These proteins were also assessed for the presence of transmembrane domains using the DAS (dense alignment surface) method (17). Proteins predicted to contain one long transmembrane helix with a strong dip in hydrophobicity in the middle of this region, corresponding to helical hairpin residues (42, 43), were designated as possessing two transmembrane domains on the basis of topology analysis of CetA (19). The G+C percentage of each HAMP-containing bipartite family member gene was plotted against the G+C percentage of the genome in which it was found. Using Excel, a linear regression trend line and associated R² value were calculated. HAMP domains and adjacent regions from homologous proteins were aligned using CLUSTALX (52) with default parameters. Conserved residues were identified using Consensus script (available at www.bork.embl-heidelberg.de/Alignment/consensus.html).

Modeling of CetA structure.

The structure of the HAMP domain from Archaeoglobus fulgidus (Protein Data Bank ID, 2ASW) was used as the foundation for modeling the structure of CetA. To model the HAMP domain of CetA, the amino acid sequence of 2ASW was virtually mutated into that of CetA by using the graphics program O (33), and the single amino acid insertion was fit using the program's lego-loop option. The resulting CetA model was then placed in a box of waters containing a minimum of two shells of water, minimized, and put through simulated annealing using torsion angle dynamics in the Crystallography & NMR system (12).

Construction of the ΔcetAB deletion mutant.

The ΔcetAB deletion mutant was constructed essentially as described by Hendrixson et al. (27). The cetA and cetB coding sequences with 1,036 bp upstream and 595 bp downstream were amplified by PCR with primers designed with KpnI sites at their 5′ ends for cloning into pUC19. The resulting plasmid was pKTY60. A deletion from the first codon of cetA to the last codon of cetB was created via Pfu mutagenesis (55). The resulting plasmid, pKTY62, was electroporated into DRH304, which harbors the cat-rpsL cassette in the cetB coding sequence. Transformants were selected on 2 mg/ml streptomycin and screened for sensitivity on 20 μg/ml chloramphenicol. The deletion was confirmed by PCR analysis and chromosomal sequencing.

Construction of a plasmid to complement the ΔcetAB mutant.

pKTY60 was digested with ApaLI and BsrBI. The resulting fragment, containing the cetA and cetB coding sequences along with 299 bases upstream and 202 bases downstream, was blunted by T4 DNA polymerase. This fragment was then cloned into the XmnI site in the E. coli/C. jejuni shuttle vector pRY108 (56).

Site-directed mutagenesis.

Point mutations in the cetA coding sequence leading to alanine substitutions (R71A, D94A, E97A, Y99A, R101A, E102A, Y116A, R117A, and K118A) were made in pKTY60 by using Pfu mutagenesis (55). The DNA sequences of the resulting plasmids were determined in order to confirm the presence of the point mutations and ensure the absence of additional mutations. These plasmids were then digested with ApaLI and BsrBI, and the resulting fragments were cloned into the XmnI site of pRY108 as described above. The orientation of the insertions into pRY108 was checked by multiple restriction digests to confirm that the resulting plasmids, pKTY361 to pKTY369, were identical to pKTY360 except for the indicated point mutations.

Conjugation of plasmids into C. jejuni.

Plasmids were conjugated into C. jejuni as described by Guerry et al. (26). Briefly, C. jejuni was grown on MH agar with 10 μg/ml trimethoprim for 16 to 20 h and resuspended in MH broth to an optical density at 600 nm (OD₆₀₀) of 1.0. Overnight cultures of the E. coli donor strain [DH5α(pRK212.1), containing the plasmid to be conjugated into C. jejuni] were diluted into fresh LB broth and grown to an OD₆₀₀ of 0.5. Five hundred microliters of the donor culture was centrifuged, and the pellet was first washed twice with MH broth and then resuspended in 1 ml of the C. jejuni recipient culture. This mixture was spotted onto MH agar with no antibiotics. After 5 h at 37°C under microaerophilic conditions, the bacteria were resuspended and spread onto MH agar containing 10 μg/ml trimethoprim, 30 μg/ml cefoperazone, 2 mg/ml streptomycin, and 50 μg/ml kanamycin. PCR was used to verify the transfer of the plasmid to the recipient C. jejuni strain.

Motility assays.

C. jejuni was first grown on MH agar containing 10 μg/ml trimethoprim and 50 μg/ml kanamycin for 16 to 20 h and then resuspended in MH broth to an OD₆₀₀ of 0.4. A 0.4-μl aliquot of each strain was injected into MH motility medium containing 0.4% agar. Plates were incubated for 28 h under microaerophilic conditions. The diameter of the outermost motility ring was measured with calipers. The average and standard deviation of six replicates were calculated, and the assay was repeated three times.

Tissue culture.

The human epithelial cell line INT 407 was used in invasion experiments. INT 407 cells were cultured in Dulbecco's modified Eagle medium (DMEM) plus 10% fetal bovine serum supplemented with GIBCO MEM nonessential amino acids and 2 mM glutamine (referred to below as DMEM) in a 37°C, 5% CO₂ incubator. When cells were cultured in the absence of C. jejuni, the DMEM was supplemented with 10 U/ml penicillin and 10 μg/ml streptomycin.

Invasion assays.

For invasion assays, INT 407 cells were seeded at approximately 10⁵/well in each well of a 24-well plate and were incubated in the absence of antibiotics for an additional 12 to 18 h. For the inoculum, C. jejuni strains were grown on MH agar for 16 to 20 h and resuspended in DMEM. The INT 407 cells were rinsed twice with phosphate-buffered saline (PBS) and inoculated with C. jejuni at a multiplicity of infection of ∼200. The 24-well plates were centrifuged at 150 × g for 5 min and then incubated in a 37°C, 5% CO₂ incubator. To determine the number of total cell-associated bacteria, the cells were incubated for 2 h, rinsed twice with PBS, and lysed in PBS plus 0.1% Triton X-100, and serial dilutions were plated onto MH agar to obtain CFU. To determine the number of intracellular bacteria, the cells were incubated for 2 h, rinsed twice with PBS, and incubated for an additional 2.5 h in DMEM plus 100 μg/ml gentamicin. The cells were then rinsed twice with PBS and lysed in PBS plus 0.1% Triton X-100, and serial dilutions were plated onto MH agar to obtain CFU. For invasion time course experiments, the cells were infected with C. jejuni as described above. At 0.5 h, 1 h, 1.5 h, 2 h, or 4 h postinfection, the numbers of total cell-associated and intracellular bacteria were determined as described above. The percentage of total cell-associated bacteria that were intracellular was calculated. The average and standard deviation of three replicates were obtained, and each invasion assay and time course assay was repeated a minimum of three times.

RESULTS

CetA and Aer differ in their predicted HAMP domain secondary structures.

Using PSI-BLAST, the CetA HAMP domain was identified as residues 47 to 101. This region was not identified as a HAMP domain by SMART analysis. This is not surprising, because there is a high degree of divergence among various HAMP domains. Current HAMP domain models miss more than 30% of HAMP homologs (B. P. Rekapalli and I. B. Zhulin, unpublished data).

The predicted structures of the HAMP domains of Aer and CetA differ substantially from one another (40). While the canonical HAMP domain consists of two amphipathic helices (AS-1 and AS-2), Aer possesses one amphipathic helix (AS-1) and one hydrophobic helix (AS-2) (Fig. 1A) (40). In contrast to the unusual secondary structure of Aer, CetA is predicted to have the more common HAMP domain structure of two amphipathic helices (Fig. 1B).

FIG. 1.

Secondary-structure analysis of the Aer and CetA HAMP domains (PSA server). The probability of each structural possibility (listed on the left) is indicated for each residue by using contour line increments of 0.1. Regions with many contour lines have a high probability of the indicated structure at that residue. Shown are Aer (A) and CetA (B) HAMP domain contour plots of secondary-structure probabilities at the indicated residues.

CetA and CetB found a new family of proposed bipartite energy taxis receptors.

Aer and CetA/CetB contain the same domains and are hypothesized to transduce energy taxis signals via similar mechanisms (27). However, the differences in HAMP domain primary and secondary structures between Aer and CetA led us to examine the CetA HAMP domain in more detail. Using the HAMP domain and proximal signaling domain (residues 44 to 139) of CetA in a BLASTP search of the nonredundant database, we identified 63 proteins with domain arrangements similar to that of CetA. Fifty-five of these proteins were encoded by genes with a neighboring ORF encoding a PAS domain and lacking other functional domains, a pattern similar to that of cetA and cetB, which are adjacent in the C. jejuni genome. We refer to these pairs of proteins encoded by adjacent genes as HAMP/PAS pairs (Table 2).

TABLE 2.

Members of the bipartite family of energy taxis transducers identified by similarity to the CetA HAMP and proximal signaling domains

Species and strain	Bacterial phylum or class	Accession no. of gene encoding the HAMP domain	E valuea	Class of HAMP-containing bipartite energy taxis receptorb	Accession no. of neighboring ORFc
Azoarcus sp. strain BH72	Betaproteobacteria	YP_934977	9.00E-04	I	YP_934978
Bradyrhizobium japonicum USDA 110	Alphaproteobacteria	NP_769616	2.00E-06	Id	NP_769615
Caminibacter mediatlanticus TB-2	Epsilonproteobacteria	ZP_01871323	0.001	I	ZP_01871324
		ZP_01870851	0.002	I	ZP_01870852
Campylobacter coli RM2228	Epsilonproteobacteria	ZP_00367208	8.00E-51	Id	ZP_00367209, ZP_00367207
Campylobacter concisus 13826	Epsilonproteobacteria	YP_001466825	2.00E-06	III	YP_001466824
Campylobacter curvus 525.92	Epsilonproteobacteria	YP_001408340e	0.001	III	YP_001408339
Campylobacter jejuni	Epsilonproteobacteria
11168		NP_282337 (cetA)	7.00E-51	Id	NP_282336 (cetB), NP_282338
81-176		YP_001000865	7.00E-51	Id	YP_001000864, YP_001000866
HB93-13		ZP_01070888	7.00E-51	I	ZP_01070815, ZP_01071201
CG8486		ZP_01809879	7.00E-51	Id	ZP_01809880, ZP_01809878
260.94		ZP_01069112	7.00E-51	Id	ZP_01069301, ZP_01069270
CF93-6		ZP_01068608	7.00E-51	Id	ZP_01068595, ZP_01068567
81116		YP_001482710	7.00E-51	Id	YP_001482711, YP_001482709
RM1221		YP_179311	7.00E-51	Id	YP_179312, YP_179310
84-25		ZP_01099629	8.00E-51	Id	ZP_01099329, ZP_01099967
Campylobacter jejuni subsp. doylei 269.97	Epsilonproteobacteria	YP_001397715	2.00E-49	Id	YP_001397716
Campylobacter lari RM2100	Epsilonproteobacteria	ZP_00368322	7.00E-34	I	ZP_00368323, ZP_00368324
Campylobacter upsaliensis RM3195	Epsilonproteobacteria	ZP_00371342	1.00E-38	I	ZP_00371343, ZP_00371341
Chromobacterium violaceum ATCC 12472	Betaproteobacteria	NP_900065	3.00E-04	I	NP_900066
Dechloromonas aromatica RCB	Betaproteobacteria	YP_287205	7	I	YP_287206
Kineococcus radiotolerans SRS30216	Actinobacteria	YP_001363132	0.009	III	YP_001363131
Magnetospirillum gryphiswaldense MSR-1	Alphaproteobacteria	CAM75236	0.003	Id	CAM75237
		CAM75032	0.005	V	CAM75033
		CAM78133	0.065	VI	CAM78134, CAM78135f
Magnetospirillum magneticum AMB-1	Alphaproteobacteria	YP_419949g	3.00E-04	I	YP_419948
		YP_423371	0.002	I	YP_423372
		YP_423064	0.06	I	YP_423063
		YP_420357	0.07	I	YP_420358
		YP_421559	0.13	I	YP_421560
		YP_419547	1.9	I	YP_419548
Magnetospirillum magnetotacticum MS-1	Alphaproteobacteria	ZP_00055894	4.00E-04	IV	ZP_00055893
		ZP_00207863	9.00E-04	I	ZP_00207862
		ZP_00208993	0.003	VIII	ZP_00208994, ZP_00051415f
		ZP_00054945h	0.43	VII	ZP_00054944
Oceanobacter sp. strain RED65	Gammaproteobacteria	ZP_01306652	0.32	I	ZP_01306653
Oceanospirillum sp. strain MED92	Gammaproteobacteria	ZP_01166142	0.051	I	ZP_01166143
Reinekea sp. strain MED297	Gammaproteobacteria	ZP_01113626	0.88	I	ZP_01113627
Rhodopseudomonas palustris	Alphaproteobacteria
BisA53		YP_783562	1.00E-04	I	YP_783563
BisB5		YP_568532	4.00E-04	I	YP_568531
		YP_571029	0.26	I	YP_571030
BisB18		YP_534502	2.00E-06	I	YP_534503
CGA009		NP_949817i	8.00E-05	I	NP_949818i
		NP_949819i	1.00E-04	II	NP_949820i
		NP_949538	1.00E-04	Id	NP_949539
		NP_949647	0.45	I	NP_949648
HaA2		YP_484693	3.00E-07	II	YP_484692
		YP_485035	8.00E-04	I	YP_485034
		YP_484934	0.16	I	YP_484933
Rhodospirillum rubrum ATCC 11170	Alphaproteobacteria	YP_428546j	0.011	I	YP_428545
Stappia aggregata IAM 12614	Alphaproteobacteria	ZP_01546232	2.00E-04	Id	ZP_01546231
Sulfurimonas denitrificans DSM 1251	Epsilonproteobacteria	YP_392559	1.00E-05	III	YP_392560
Wolinella succinogenes DSM 1740	Epsilonproteobacteria	NP_907800	1.00E-11	I	NP_907801
		NP_906923	6.00E-09	Id	NP_906922
		NP_907510	3.00E-05	Id	NP_907511

^aObtained from a BLAST search with the CetA HAMP and proximal signaling domains (amino acids 44 to 139).

^bSee Fig. 2.

^cThe neighboring ORF encodes a PAS domain, but no other domains unless so indicated. Where two accession numbers are given, the gene encoding the HAMP-containing protein is flanked by two genes encoding PAS domain proteins.

^dTransmembrane helix prediction programs predict one transmembrane helix, but we propose that there are two transmembrane helices present in a helical hairpin (see Materials and Methods).

^eMisannotated as a DNA binding response regulator.

^fThe neighboring ORF contains additional predicted functional domains.

^gMisannotated as a sensory rhodopsin II transducer.

^hMisannotated as an MCP but does not contain the highly conserved domain.

ⁱThese HAMP/PAS pairs are adjacent to one another.

^jMisannotated as a response regulator.

The CetA-like (HAMP-containing) members of this family can be divided into eight classes on the basis of predicted topology and functional domains (Fig. 2). Most CetA homologs fall into classes I to III, which contain a HAMP domain, a signaling domain, and two, one, or no transmembrane domains, respectively. Only two proteins contain a single predicted transmembrane domain. Single pass transmembrane proteins are unusual in bacteria but have been identified (16, 41). Further examination is necessary, however, to ascertain whether this topology prediction holds true or whether, perhaps, the predicted start site of translation is incorrect. Variations on the domain arrangement of CetA, with proteins containing additional domains or lacking the signaling and/or transmembrane domain, were also identified (classes IV to VIII). In all, we identified 55 HAMP/PAS pairs, 40 of which are in genera other than Campylobacter (Table 2). The majority of organisms containing HAMP/PAS pairs belong to the epsilon class of Proteobacteria, but HAMP/PAS pairs are also found in the alpha, beta, and gamma classes of Proteobacteria as well as in one gram-positive organism, the actinobacterium Kineococcus radiotolerans (Table 2). The GC content of the HAMP genes correspond well to those of the genomes in which they are found (Fig. 3). Several strains contain more than one HAMP/PAS pair. In those strains that contain multiple HAMP/PAS pairs, the HAMP members of different HAMP/PAS pairs may also fall into more than one of the classes defined in Fig. 2. We propose that many of these HAMP/PAS pairs identified by similarity to the CetA HAMP domain comprise a new family of bipartite energy taxis receptors.

FIG. 2.

Classes of HAMP-containing bipartite family members. HAMP-containing proteins from Table 2 were analyzed by the SMART and DAS tools and separated into different classes on the basis of predicted topology and functional domains. SD indicates the signaling domain conserved in MCPs. PilZ domains are cyclic di-GMP effector domains.

FIG. 3.

The G+C percentage of each gene encoding a HAMP protein in the bipartite family was plotted against the G+C percentage of the genome in which it resides. A linear regression trend line and R² value are shown. The data point for the Kineococcus radiotolerans HAMP protein is circled. An arrow indicates the data point for the Magnetospirillum magneticum AMB-1 HAMP protein found in the magnetosome island (see Discussion).

Bipartite family members contain conserved HAMP and proximal domain residues.

The HAMP domain and proximal signaling domain of representative CetA-like bipartite family members were aligned using CLUSTALX (Fig. 4A). Eleven residues are highly conserved within this subfamily but not within the HAMP consensus sequence. HAMP domains, in general, share very low levels of sequence conservation, so the presence of such highly conserved residues is notable. Most of these highly conserved residues are either charged or aromatic, consistent with the possibility that they could play roles in protein-protein interactions.

FIG. 4.

Conserved residues within the HAMP domain and a proximal connector to the signaling domain of the HAMP-containing bipartite family members. This multiple alignment of representative CetA homologs includes species from different divisions of Proteobacteria and the actinobacterium. Both representative orthologs and paralogs are shown. The boundaries of the predicted HAMP domain and the N-terminal start of the signaling domain are shown above the alignment. We refer to the region between the HAMP and signaling domains as a connector. Each sequence is identified by its NCBI locus tag (e.g., Cj1190c is the locus tag for the CetA protein). Initial characters in locus tags identify species and/or strain names: Cj, Campylobacter jejuni NCTC 11168; CMTB2, Caminibacter mediatlanticus TB-2; WS, Wolinella succinogenes DSM 1740; azo, Azoarcus sp. strain BH72; MED92, Oceanospirillum sp. strain MED92; MED297, Reinekea sp. strain MED297; CV, Chromobacterium violaceum ATCC 12472; Daro, Dechloromonas aromatica RCB; RPB, Rhodopseudomonas palustris HaA2; blr, Bradyrhizobium japonicum USDA 110; amb, Magnetospirillum magneticum AMB-1; Rru, Rhodospirillum rubrum ATCC 11170; SIAM614, Stappia aggregata IAM 12614; Krad, Kineococcus radiotolerans SRS30216. Strongly conserved positions likely to contribute to protein-protein interactions are highlighted in blue (negative charge), red (positive charge), or green (aromatic). Strongly conserved positions likely to contribute primarily to structure and mildly conserved positions are highlighted in gray. The conservation consensus (cons.) at 100% and 80% (calculated using the Consensus script, available at http://coot.embl.de/Alignment//consensus.html) is shown below the alignment. The types of residues are abbreviated as follows: h, hydrophobic; t, turn-like; p, polar; s, small; l, aliphatic; c, charged; a, aromatic. Residues that are conserved at 100% make up the signature of the entire bipartite family (conserved at 90% within the entire family).

To understand the placement of these conserved residues in the protein, a model of the CetA HAMP domain was created. The amino acid sequence of the CetA HAMP domain was threaded onto the averaged nuclear magnetic resonance structure of Archaeoglobus fulgidus (2ASW), the only HAMP domain present in the Protein Data Bank, using the graphics program O (33) and the alignment shown in Fig. 5A. Four of the five conserved residues that fall within the HAMP domain (R71, D94, E97, and R101) appear to cluster near one another on the surface of the HAMP four-helix bundle (Fig. 5B). The fifth conserved residue in the HAMP domain (Y99), however, is located between the two monomers and is predicted to be involved in tyrosine stacking interactions between HAMP monomers at the base of the four-helix bundle (Fig. 5B). The six remaining conserved residues (E102, F115, Y116, R117, K118, and F128) fall outside of the Af1503 structure used to model the CetA HAMP domain, and therefore, their location within the protein cannot be predicted. This model of the CetA HAMP domain structure shows an overall dipole moment, as the N-terminal half of the structure has a net positive charge, while the C-terminal half of the structure has a net negative charge (Fig. 5C).

FIG. 5.

Model of CetA. (A) The sequence alignment of the CetA HAMP domain and Af1503 from Clustal W2 (15) (http://www.ebi.ac.uk/Tools/clustalw2/index.html) is shown with the hydrophobic residues of the heptad repeat and linker region boxed in yellow. The secondary-structure prediction for CetA (18) (http://www.compbio.dundee.ac.uk/∼www-jpred/) and the published secondary structure of Af1503 are shown above and below their amino acid sequences, respectively. Helices and β-strands are represented as solid rectangles and arrows, respectively. Residues mutated in CetA are italicized. This alignment was used in the creation of the structural model of the CetA HAMP domain shown in panels B and C. (B) The model of CetA is depicted as a ribbon diagram, with the mutated residues shown as sticks. The views for individual molecules are separated by a 90° rotation about the y axis. (C) Electrostatic surface potentials for the modeled structure of CetA were calculated using APBS (7) and mapped onto their respective solvent-accessible surfaces by using the PyMOL molecular graphics system (http://www.pymol.org/). Negative potentials (−10 kT/e) are shown in red, positive potentials (10 kT/e) in blue. The views are the same as in panel B. The protein structures are shown at the same magnification for each view.

Conserved HAMP domain residues are important for CetA function.

Attempts to complement a ΔcetA mutant with a plasmid expressing cetA from a constitutive promoter were unsuccessful (data not shown), likely because the relative levels of CetA and CetB, as well as the levels of these proteins relative to other MCPs, are important for proper energy taxis signal transduction. A double-deletion strain was constructed with the deletion extending from the first codon of cetA to the last codon of cetB. We also constructed a plasmid, pKTY360, consisting of the cetA and cetB genes, as well as 299 bases upstream and 202 bases downstream, cloned into the pRY108 E. coli/C. jejuni shuttle vector (56). cetAB expression in pKTY360 presumably originates from the native promoter, since there is no promoter to drive the expression of cloned DNA in pRY108. pKTY360 complements the motility defect of the ΔcetAB double mutant in MH motility agar but has no effect on the motility defect of the ΔrpoN mutant, which is nonmotile and lacks flagella (Fig. 6).

FIG. 6.

Influence of point mutations on the function of CetA in motility. Motility assays were performed on wild-type (wt), ΔcetAB, and ΔrpoN strains containing an empty vector, pRY108, or pRY108::cetAB (with or without the indicated HAMP and proximal signaling domain point mutations).

Alanine substitutions were made for nine of the conserved residues in CetA in the context of pKTY360. Most of these substitutions led to a growth defect (D94A, E97A, R101A, E102A, and R117A) or loss of CetA stability (Y116A) (data not shown). Three alanine substitution mutants (R71A, Y99A, and K118A [pKTY361, pKTY364, and pKTY369, respectively]) retained growth kinetics and CetA expression levels at or near those for the ΔcetAB mutant complemented with pKTY360 (data not shown). Mutant proteins with these substitutions exhibited an abrogated or reduced ability to rescue the motility defect of the ΔcetAB double-deletion strain (Fig. 6). The motility pattern of the HAMP mutants on MH agar may also be altered in shape or ring structure, possibly reflecting CetA/CetB sensing. However, unlike the media used for E. coli motility plates, the MH motility medium used for C. jejuni is too opaque for such distinctions to be observed. Still, these results indicate that at least some of the conserved residues within the bipartite family HAMP and proximal signaling domains are required for wild-type function.

The ΔcetA mutant has an epithelial cell invasion defect, but the ΔcetB mutant does not.

CetA and CetB are each required for energy taxis in C. jejuni, suggesting a functional interaction whose mechanism may be similar to that of Aer (27). However, the separation of Aer domains into two proteins in CetA and CetB, as well as in other members of this family, raises the possibility that each member of a HAMP/PAS pair may contribute independently to different traits. Because C. jejuni is a common commensal of chickens, we previously tested ΔcetA and ΔcetB mutants in a chick colonization model, and both mutants colonized to wild-type levels (28).

C. jejuni actively invades human epithelial cells, a trait associated with its pathogenicity (37). To determine whether or not CetA or CetB contributes to this phenotype, we tested mutants in a tissue culture model of invasion (29, 44). INT 407 cells were infected with the wild type (DRH212) or with a ΔcetA, ΔcetB, or ΔcetAB mutant at a multiplicity of infection of ∼200. Immediately upon infection, bacteria were centrifuged onto the INT 407 cells in order to rule out any effects of motility on the invasion assay. After 2 h, the number of total cell-associated bacteria was determined in half of the wells (see Materials and Methods), and gentamicin was added to the remaining wells to kill extracellular bacteria. After a further 2.5 h, the number of intracellular bacteria was determined as described in Materials and Methods, and the percentage of total cell-associated bacteria that were intracellular (i.e., that had invaded) was calculated. Strains lacking cetA alone, or lacking both cetA and cetB, invaded INT 407 cells approximately 5 times less efficiently than the wild type, whereas a ΔcetB mutant invaded at wild-type levels (Fig. 7A). C. jejuni does not grow significantly in DMEM or intracellularly over the time period of these invasion assays, and there were no differences in survival between strains during the course of these experiments (K. T. Elliott and V. J. DiRita, unpublished data).

FIG. 7.

Effects of cetA and cetB mutations on epithelial cell invasion. (A) INT 407 cells were infected with the wild type (wt) or the ΔcetB, ΔcetA, or ΔcetAB mutant. The percentages of total cell-associated bacteria that were intracellular following a 2-h infection were calculated. (B) INT 407 cells were infected with the wt or the ΔcetA mutant. The percentages of total cell-associated bacteria that were intracellular following a 30-min, 1-h, 1.5-h, 2-h, or 4-h infection are shown.

The invasion defect exhibited by the ΔcetA mutant was investigated further by analyzing the kinetics of invasion. The percentage of total cell-associated bacteria that had invaded the INT 407 cells was determined at various times between 30 min and 4 h postinfection (Fig. 7B). The level of invasion by the ΔcetA mutant remained lower than that of the wild type at all times. The rate of invasion by the ΔcetA mutant was initially much lower than that by the wild type. By 2 to 4 h postinfection, however, the rate of invasion by the ΔcetA mutant reached near-wild-type levels. These results indicate that the ΔcetA mutant lags behind the wild type in the initiation of invasion.

DISCUSSION

We identified a new family of apparent bipartite energy taxis receptors that contain HAMP and proximal signaling domains homologous to CetA. These domains have several highly conserved residues, at least three of which are required for CetA function in C. jejuni motility. We propose that these conserved residues, several of which cluster on the exterior of the base of the HAMP domain four-helix bundle, make up the PAS interaction surface.

In further studies of the CetA/CetB HAMP/PAS pair, we demonstrated that CetA is required for the invasion of human epithelial cells by C. jejuni while CetB is dispensable. These findings support the hypothesis that CetA (and perhaps other HAMP-containing members of the bipartite family) can act independently of its PAS partner to regulate traits other than energy taxis. CetA could control invasion either alone or through interactions with as yet unknown proteins (Fig. 8). Similarly, CetB could perhaps regulate traits independently of CetA, although we have yet to identify a CetB-dependent phenotype other than energy taxis. Thus, this work identified a new family of apparent bipartite energy taxis receptors and provided evidence of a functional consequence of having the domains of Aer separated into distinct proteins within this family.

FIG. 8.

Proposed model for CetA/CetB function compared to Aer function. CetA and CetB are proposed to transduce an energy taxis signal via a mechanism similar to that of Aer. However, CetA is proposed to interact with another, unidentified protein to promote invasion, while CetB does not. See Discussion for details.

Unique features of the CetA HAMP domain suggest that it functions differently from that of Aer.

The HAMP domains of Aer and CetA differ substantially from one another. Aer has a hydrophobic helix (AS-2) where most HAMP domains have a second amphipathic helix (Fig. 1A) (40). The hydrophobic nature of the Aer AS-2 is proposed to reflect the fact that this helix is involved both in HAMP-HAMP interactions (between Aer monomers in a dimer), and in HAMP-PAS interactions (40). Since Aer and CetA/CetB are proposed to transduce an energy taxis signal via similar mechanisms, we predict that the HAMP domain of CetA would also be involved in both HAMP-HAMP and HAMP-PAS interactions. If so, then the fact that the CetA HAMP domain, in contrast to that of Aer, contains two amphipathic helices (Fig. 1B) is unexpected. Assuming that the CetA AS-2 interacts directly with CetB, the molecular nature of this interaction must differ substantially from that of the interaction between the HAMP and PAS domains of Aer.

A newly identified family of HAMP/PAS protein pairs.

Based on similarity with the CetA HAMP and proximal signaling domains and genome context, numerous HAMP/PAS pairs homologous to CetA and CetB were identified. We hypothesize that most of the members of this protein family function as bipartite energy taxis receptors. Based on their bipartite nature and on the alternative domain architectures present within this family (Fig. 2), some pairs may have other functions in addition to, or instead of, energy taxis (see below). The species containing HAMP/PAS pairs represent a broad range of bacteria—alphaproteobacteria, betaproteobacteria, epsilonproteobacteria, gammaproteobacteria, and one gram-positive actinobacterium (Table 2)—including human pathogens, animal commensals, plant symbionts, and species found in marine, aquatic, and soil environments. Together, these observations suggest that these HAMP/PAS pairs have been conserved under a diverse set of selection pressures. Our analysis of the GC contents of the HAMP-containing bipartite family genes suggests that they were not spread recently by horizontal gene transfer. There are a few possible exceptions to this, with slightly lower GC contents in the HAMP gene than in the overall genome. One of these (Fig. 3) occurs in a known genomic island associated with magnetoaerotaxis (22). GC content comparisons, however, represent only one criterion for gene acquisition by horizontal transfer. Further bioinformatic evaluation of the bipartite family genes and the genomes in which they reside is necessary in order to more definitively ascertain whether these genes have been propagated by such a mechanism.

Other than CetA and CetB, none of the HAMP/PAS pairs has been studied beyond sequence analysis. However, the various classes of CetA-like HAMP-containing proteins present some novel functional possibilities (Fig. 2). Class IV has a domain that has been implicated in osmotic stress response, resistance to reactive oxygen species, and redox regulation (3, 32, 46). Class V proteins in this family have a PilZ domain, thought to be a cyclic diguanylate effector domain (8). Class VI has the same domains as CetA, but with a larger periplasmic loop, perhaps enabling it to sense an extracellular signal in addition to the signal that originates in the PAS domain. We predict that each of these classes possesses increased functional flexibility, with the additional domains noted above providing alternative means of input to or output from the HAMP domain.

In several Campylobacter species, there are two PAS neighbors encoded by genes flanking the gene encoding the HAMP protein. These are likely homologous to cetB and cj1191c, which flank cetA in the C. jejuni 11168 genome (27). The role of cj1191c remains unclear. Multiple attempts at constructing an in-frame deletion in this ORF have been unsuccessful (Elliott and DiRita, unpublished data). Whether or not the PAS domains of each of these proteins interact with their HAMP neighbors is unknown.

Conserved HAMP domain residues are required for CetA function.

Several highly conserved residues that are not conserved in a canonical HAMP domain are found in the HAMP and proximal signaling domains of the bipartite family members (Fig. 4A). These are in the connector between AS-1 and AS-2, in AS-2, and in the proximal signaling region. Alanine substitutions at nine of these positions in CetA were investigated, and only three of the point mutations (R71A, Y99A, and K118A) enabled growth kinetics and CetA expression levels comparable to those of wild-type cells. These three mutant proteins were unable to restore motility when expressed in the ΔcetAB mutant (Fig. 6), prompting our conclusion that these residues are required for wild-type function of CetA in motility. Complementation tests with these mutant plasmids for the invasion phenotype of the ΔcetA mutant have been difficult to interpret, perhaps due to expression of the complementing alleles on a multiple-copy replicon (Elliott and DiRita, data not shown). Testing of these residues for their role in C. jejuni invasion may require that we place the mutant alleles on the chromosome, a process that works with varying efficiency in C. jejuni. We cannot draw any conclusions about the role of the other six conserved residues in CetA function. Other substitutions may be required in order to probe the role of these residues.

Structural implications for signaling transduction through CetA and CetB.

The current model of signal transduction through Aer includes a direct interaction between the PAS and HAMP domains, allowing the FAD redox signal to be transmitted parallel to the inner membrane (50). Current evidence for such an interaction is thus far indirect. Deletion of, or mutations in, the Aer HAMP domain disrupt Aer maturation and FAD binding (9, 13, 40). A point mutation in the HAMP domain that abrogated aerotaxis by Aer could be specifically suppressed by a second-site mutation in the PAS domain (54). We propose that the CetA HAMP domain and CetB interact, similarly to the proposed PAS-HAMP interaction in Aer, and studies to investigate this proposed interaction are currently in progress. Further, we predict that the other HAMP/PAS pairs in the bipartite family identified in this study also interact with one another.

Modeling the structure of the Aer HAMP domain onto that previously determined for Af1503 led to the observation that point mutations resulting in a constant “signal-on” state of Aer cluster together at the base of the four-helix bundle formed by the HAMP dimer (54). These point mutations may strengthen the proposed PAS-HAMP interaction and define the PAS-HAMP interaction surface (54). Most of the conserved residues within the HAMP domains of the bipartite family members are predicted to be located at the base of the HAMP dimer four-helix bundle, near one another on the surface of this dimer (Fig. 5B). We hypothesize that this region of the HAMP domain dimer plays a role in HAMP-PAS interactions between members of the HAMP/PAS pairs in this family of proteins.

The one residue in the CetA HAMP domain model that does not fall on the same surface as the others is Y99. While the role of this tyrosine in HAMP domain function remains unclear, the location of these tyrosines in our model lends some support to the model of HAMP domain signaling proposed by Hulko et al. (31). Specifically, if the helices of the HAMP domain four-helix bundle do rotate relative to one another, they must do so in the direction proposed by Hulko et al. (31); rotation in the opposite direction is prevented by steric hindrance due to the location of Y99.

Our model suggests that the HAMP domain of CetA is a polar structure, with a positively charged N terminus and a negatively charged C terminus (Fig. 5C). This dipole moment is apparent, but less pronounced, in the Af1503 structure (data not shown) and may be a previously unrecognized feature of HAMP domains. Since the N termini of HAMP domains are generally proximal to the inner membrane, it may be that the net positive charge in this region acts as an attractive force, further tethering the HAMP domain to the membrane. The role of the net negative charge of the base of the HAMP domain is more speculative. Since this region has been implicated in the PAS-HAMP interaction in Aer, we propose that there may be a cognate positive surface on the CetB PAS domain facilitating interaction between CetA and CetB. Since the AS-2 helix of Aer is hydrophobic, we would predict that the equivalent surface of the Aer PAS domain would comprise a hydrophobic patch.

CetA, but not CetB, contributes to cell invasion.

C. jejuni is one of the most prevalent causes of bacterial gastroenteritis in the United States (21). While not considered a highly invasive organism compared to bacteria such as Salmonella and Shigella spp., C. jejuni actively invades nonphagocytic human epithelial cells in tissue culture models (34, 44). Our studies indicate that the ΔcetA mutant and the ΔcetAB mutant have an approximately fivefold defect in invasion compared to the wild type (Fig. 7A). Compared to those of some other known C. jejuni invasion mutants, the magnitude of the invasion defect of the ΔcetA mutant is relatively small (24, 36, 49). What is striking about our observations regarding invasion is not the magnitude of the ΔcetA effect but the fact that a cetB mutation shows no defect (Fig. 7A). If CetA and CetB function solely as partners to transduce an energy taxis signal, we would expect the ΔcetA and ΔcetB mutants to have similar phenotypes. It should be noted that CetB levels are quite low in the ΔcetA mutant (19). However, the lack of an invasion defect in the ΔcetB mutant directly rules out a role for CetB in invasion. We conclude that CetA and CetB function independently of one another to regulate invasion. A previous study found that a transposon insertion in cetB resulted in increased adherence to INT 407 cells and decreased invasion of INT 407 cells (23). These results clearly differ from our findings that the ΔcetB mutant is not affected in adherence (Elliott and DiRita, data not shown) or invasion. The source of the discrepancy between those findings and ours is not clear, although that study used a transposon insertion mutant, as opposed to the in-frame deletion used in our study.

One explanation for the invasion defect of the ΔcetA mutant is that it results from the motility defect of this mutant. To reduce the contribution of motility to invasion in these experiments, the bacteria were brought into contact with the INT 407 cells by low-speed centrifugation at the beginning of the assay. Further, we eliminated this potential contribution of motility to our results by determining the percentage of total cell-associated bacteria that were intracellular. By this method of calculation, the ΔcetA and ΔcetAB mutants still have an approximately fivefold defect in invasion (Fig. 7A). Given these considerations, we conclude that the invasion defect of the ΔcetA mutant cannot be attributed solely to the motility defect.

The invasion defect of the ΔcetA mutant results from an initial lag in the rate of invasion. By 2 to 4 h postinfection, the ΔcetA mutant invades at rates near or at that of the wild type (Fig. 7B). The mechanism of C. jejuni invasion is still being dissected but appears to be mainly microtubule dependent (37), a feature that differentiates the C. jejuni invasion process from those of many invasive pathogens that use host cell actin for internalization. Proteins secreted from the flagellum are also required for C. jejuni invasion (35, 36, 48). Metabolic labeling experiments showed no significant changes in the ability of the ΔcetA mutant to secrete proteins associated with invasion (S. A. Pacheco, M. E. Konkel, K. T. Elliott, and V. J. DiRita, unpublished data). Additionally, C. jejuni invasion involves host cell protein kinase activity as well as a subset of small Rho GTPases (11, 30, 38), and we speculate that the ΔcetA mutant has a defect in the ability to initiate these signaling events. A recent study observed that C. jejuni can migrate beneath (“subvade”) epithelial cells in tissue culture prior to invasion (53). However, it was observed that increased subvasion efficiency correlated with decreased CheW expression and decreased motility in soft agar (53). Therefore, if CetA contributes to subvasion, it would appear that it must do so independently of CetB and the chemotactic machinery. Until more mechanistic details about both subvasion and the initiation of epithelial cell signaling events are known, the molecular mechanisms behind the invasion defect of the ΔcetA mutants will remain to be elucidated.

In summary, we identified a new family of proposed bipartite energy taxis receptors, similar to the CetA/CetB system in C. jejuni. Although we suggest that the HAMP/PAS pairs in this family transduce an energy taxis signal via a mechanism similar to that of Aer, there are clear departures from the Aer model. Differences between CetA and Aer in their predicted HAMP domains, as well as the presence of highly conserved residues within the HAMP and proximal domains of the CetA family members, suggest that the nature of the PAS-HAMP interaction within this family is mechanistically different from that in Aer. Finally, the involvement of CetA, but not CetB, in epithelial cell invasion supports our hypothesis that the members of HAMP/PAS pairs may act independently of each other to control phenotypes other than energy taxis.