Transcriptional Regulation of the tad Locus in Aggregatibacter actinomycetemcomitans: a Termination Cascade

Abstract

The tad (tight adherence) locus of Aggregatibacter actinomycetemcomitans includes genes for the biogenesis of Flp pili, which are necessary for bacterial adhesion to surfaces, biofilm formation, and pathogenesis. Although studies have elucidated the functions of some of the Tad proteins, little is known about the regulation of the tad locus in A. actinomycetemcomitans. A promoter upstream of the tad locus was previously identified and shown to function in Escherichia coli. Using a specially constructed reporter plasmid, we show here that this promoter (tadp) functions in A. actinomycetemcomitans. To study expression of the pilin gene (flp-1) relative to that of tad secretion complex genes, we used Northern hybridization analysis and a lacZ reporter assay. We identified three terminators, two of which (T1 and T2) can explain flp-1 mRNA abundance, while the third (T3) is at the end of the locus. T1 and T3 have the appearance and behavior of intrinsic terminators, while T2 has a different structure and is inhibited by bicyclomycin, indicating that T2 is probably Rho dependent. To help achieve the appropriate stoichiometry of the Tad proteins, we show that a transcriptional-termination cascade is important to the proper expression of the tad genes. These data indicate a previously unreported mechanism of regulation in A. actinomycetemcomitans and lead to a more complete understanding of its Flp pilus biogenesis.

Known primarily as the causative agent of localized aggressive periodontitis, the bacterium Aggregatibacter (formerly Actinobacillus) actinomycetemcomitans is a facultatively anaerobic, gram-negative coccobacillus found in the oral cavities of humans and Old World monkeys (16, 27, 30, 51). A. actinomycetemcomitans is also associated with systemic infections, such as infective endocarditis, septicemia, and abscesses (10, 46). A characteristic of fresh clinical isolates is their extremely tenacious adherence to a wide variety of surfaces (11). The wild-type cells display several phenotypes: rough-colony formation, autoaggregation, pilus biogenesis, and biofilm formation (11, 16, 21). The phenotypes are correlated with the production of long, bundled Flp pili, termed Flp fibers (20). Upon laboratory subculture, these isolates can undergo a transition to a nonadherent and nonpiliated form (12). The tad (tight adherence) locus is necessary for this adherence and the other Flp-related phenotypes. Using a rat model of localized aggressive periodontitis, we have demonstrated the necessity of the tad genes for colonization and the oral pathogenesis of A. actinomycetemcomitans (20, 40, 45).

The tad locus of A. actinomycetemcomitans includes 14 genes (flp-1-flp-2-tadV-rcpCAB-tadZABCDEFG) (Fig. 1), 12 and probably 13 of which are essential for Flp pilus biogenesis (20, 22, 33, 35). We have shown that tad-like genes are widespread in both the eubacterial and archaeal kingdoms (35, 37, 45). At least one copy of the tad locus has been found in many bacterial genomes, including nonpathogenic species, as well as important human pathogens (45). The genes for some of the proteins encoded by the tad locus have homology to genes for other bacterial secretion proteins and can be assigned functions based on sequence analysis and experimental data. flp-1 encodes a type IVb prepilin (22), which is proteolytically processed by TadV, as are the TadE and TadF pseudopilins (43). Processed Flp1 is the major component of the Flp pili (17) and is posttranslationally modified, probably by glycosylation (17, 43). The tad locus in A. actinomycetemcomitans also encodes the RcpA outer membrane secretin (7), the TadA ATPase (4, 34), and the PilC-like TadB and TadC proteins (32). Their genes show homologies to type II and type IV secretion or type IV pilus genes, and their protein products have known or strongly predicted functions (18, 26, 35). The genes of other proteins (RcpB, RcpC, TadD, TadE, TadF, TadG, and TadZ) do not show extensive homologies to any genes and cannot be assigned predicted functions.

FIG. 1.

The tad locus of A. actinomycetemcomitans. The position and direction of transcription are indicated for each gene.

Little is known about the transcription of the tad locus and its possible regulation. Others have shown that the intergenic regions of the genes between flp-1 and tadD are transcribed, leaving open the possibility that the tad locus genes are transcribed as a single unit (15). A transcriptional start site and a putative, strong promoter that appear 102 bp and 142 bp upstream of the first gene of the tad locus, respectively, have been identified (15). The promoter was shown to be sufficient for transcription of a reporter gene in Escherichia coli (15). Several spontaneous nonadherent and nonpiliated (smooth) strains of A. actinomycetemcomitans contain a mutation in the identified promoter (48).

Genes that encode macromolecular complexes are sometimes arranged in polycistronic units (9). The arrangement of the tad locus of A. actinomycetemcomitans hints at a single polycistronic message, but some locus proteins are needed in different amounts; e.g., many copies of the Flp1 protein compose the Flp pilus (17), while TadA, an ATPase, is probably needed in six copies per complex (4). Other Tad proteins are likely to be present in even smaller amounts.

In this study, we show that the promoter previously identified (15) is both necessary and sufficient for transcription in A. actinomycetemcomitans and for expression of at least flp-1. We also identify a transcriptional termination cascade that likely governs the stoichiometry of some of the proteins of the tad locus. Two terminators appear to be Rho independent, and one is likely to be Rho dependent. We predict that this termination cascade regulates the relative levels of tad genes and therefore the correct stoichiometry of the Tad proteins, leading to efficient pilus biogenesis.

MATERIALS AND METHODS

Bacterial strains, growth media, and plasmids.

The bacterial strains and plasmids used in this study are listed in Table 1. CU1000N, a nalidixic acid-resistant derivative of a rough clinical isolate, was used as the wild-type strain (20). CU1060N is a nalidixic acid-resistant, spontaneous smooth (nonadherent) mutant of CU1000 (12). The A. actinomycetemcomitans strains were grown in A. actinomycetemcomitans growth medium (AAGM), as previously described (20). As necessary, spectinomycin (20 μg/ml) and/or nalidixic acid (20 μg/ml) was used as a supplement.

TABLE 1.

Bacterial strains and plasmids

Strain or plasmid	Relevant characteristicsa	Reference or source
A. actinomycetemcomitans
CU1000N	Spontaneous nalidixic acid-resistant mutant of rough clinical isolate CU1000, serotype f; Nalr	20
CU1060N	Spontaneous smooth mutant of CU1000 (CU1060) with mutation in −10 position of tadp; spontaneous nalidixic acid-resistant mutant of CU1060; Nalr	12
E. coli
TOP10	F−mcrA Δ(mrr-hsdRMS-mcrBC) ΔlacX74 deoR recA1 araD139 Δ(ara-leu)7697 galU galK rpsL (Strr) endA1 nupG (φ80lacZΔM15)	Invitrogen
Plasmids
pRK21761	RK2 IncP Kmr Apr Tcs Lac+ (tetA::lacZ) oriT1 Mob+	41
pJAKlac	Promoterless lacZ gene in pJAK13; Spr Δ(tacp lacIq)	This study
pKK008	tacp promoter with the R1− region in pJAKlac	This study
pKK009	tacp promoter with the R1+ region in pJAKlac	This study
pKK010	tacp promoter with the R3− region in pJAKlac	This study
pKK012	tacp promoter with the R2− region in pJAKlac	This study
pKK013	tacp promoter with the R2′ region in pJAKlac	This study
pKK014	tacp promoter with the R2+ region in pJAKlac	This study
pKK015	tacp promoter with the R3+ region in pJAKlac	This study

^aAp^r, ampicillin resistance; Km^r, kanamycin resistance; Nal^r, nalidixic acid resistance; Sp^r, spectinomycin resistance; Str^r, streptomycin resistance, Tc^s, tetracycline sensitive.

E. coli TOP10 (Table 1 shows the genotype) (Invitrogen) was used as the host strain for the IncP mobilization plasmid (pRK21761) and mobilizable IncQ plasmids, which were introduced into A. actinomycetemcomitans by conjugation (13). Briefly, the poorly self-transmissible oriT-defective, but mobilization-proficient, pRK21761 was used to mobilize the IncQ plasmids. E. coli was grown on Luria-Bertani (LB) agar plates or in LB broth at 37°C (38). Kanamycin (50 μg/ml) and/or spectinomycin (50 μg/ml) was used for appropriate selection of plasmids.

DNA manipulations.

Genomic DNA from A. actinomycetemcomitans was extracted using the DNeasy tissue kit (Qiagen). For cloning and probe construction, ExTaq (TaKaRa) was used for PCR amplification. The primers, synthesized by Invitrogen, are listed in Table 2. PCR products for cloning were digested with BamHI and XbaI before being ligated to the plasmid vector using T4 DNA ligase (New England Biolabs). All restriction enzymes were from New England Biolabs. Transformation of recombinant plasmids into E. coli Top10 was done as described previously (8). Plasmid DNA isolation was done using the Qiaprep Spin miniprep kit (Qiagen). Agarose gel electrophoresis was used for isolation of DNA probes, which were purified using the Qiaex II gel extraction kit (Qiagen). Sequencing of DNA was done using the plasmid-specific primer lacZseq67rev (Table 2).

TABLE 2.

Oligonucleotides used in this study

Primer name	Sequence (5′→3′)a	Nucleotideb	Description
Northern probes
flp-1-5RT	ACTACTAAAGCATACATCAAAGC	45	flp-1 5′ end
flp-1-3RT	TTTGCACTTGCAACTGTACTAG	223	flp-1 3′ end
flp-2-5RT	GGATTTACTGGATTACTTTTATC	360	flp-2 5′ end
flp-2-3RT	TTAGCTACTTTTGCTAACTATAG	563	flp-2 3′ end
tadV-5RT	ATGAACTGGGTTATTAATGCC	599	tadV 5′ end
tadV-3RT	CCGAAACAAGCTTTACATCTC	825	tadV 3′ end
Constructs
flp-79 XbaIFw	TGTCTAGACAAAATGCATACACATATATAGCG	−181	5′ end of tadp region
flp-28 BamHIRev	CGGGATCCCAGAATTACGAGGATAACAACAAATATC	−1	3′ end of tadp region up to translational start of flp-1
flp-100 BamHIRev	CGGGATCCCGAGTGGCATTATAGTACTG	−102	3′ end of tadp region not including putative transcriptional start
FlpXstart BamHIRv	CGGGATCCTTTTAACGAGTGGCATTATAGTAC	−100	3′ end of tadp region including just transcriptional start
tadP5flp1Xba	CGTCTAGATTATTCTAAAAAACTTGCATTTTAATTTTTCAGTACTATAATGCCACTCGTCTGTTGCTGTATAATCGTTGC	142	5′ end of R1 with tadp promoter and XbaI site
tacP5flp1Xba	CGTCTAGATGTTGACAATTAATCATCGGCTCGTATAATGTGTGGAATTGTGAGCGGATAACAATTTCACCTGTTGCTGTATAATCGTTGC	142	5′ end of R1 with tacp promoter and XbaI site
3SLBam	GCGGATCCCAAAATATAGCAGTAGATAGCC	324	3′ end of R1+ with BamHI site
3NoSLBam	GCGGATCCTTAATATTTAAGTTGTTATTTATTACT	254	3′ end of R1− with BamHI site
tacP5tadVXba	CGTCTAGATGTTGACAATTAATCATCGGCTCGTATAATGTGTGGAATTGTGAGCGGATAACAATTTCACGATTGGGACTAATTATTATCGG	909	5′ end of R2 with tacP promoter and XbaI site
3NotadVBam	GCGGATCCGTTTGATAGAGCTAGGTTTATC	1003	3′ end of R2− with BamHI site
31tadVBam	GCGGATCCCTGCTCATCAAGACAATATTAG	1024	3′ end of R2′ with BamHI site
32tadVBam	GCGGATCCCCCTTAATTAAATTTTAACAAATTAAAGU	1049	3′ end of R2+ with BamHI site
tacP5tadGXba	CGTCTAGATGTTGACAATTAATCATCGGCTCGTATAATGTGTGGAATTGTGAGCGGATAACAATTTCACCTGTAAACAGTAAGCAAGGC	11492	5′ end of R3 with tacp promoter and XbaI site
3NotadGBam	GCGGATCCTATTGGAACAATTTCGGTTTTTG	11573	3′ end of R3− with BamHI site
3SLtadGBam	GCGGATCCGATGAGGTCATTTTTTTGAGC	11665	3′ end of R3+ with BamHI site
Sequencing
lacZseq67rev	GTAACGCCAGGGTTTTCCC		For sequencing from 5′ end of lacZ through inserts

^aAdded restriction enzyme sites are in italics; added promoter sequences are in boldface.

^bNucleotide relative to flp-1 translational start site (21).

Reporter construction.

The reporter plasmid pJAKlac was constructed as a transcriptional fusion using the backbone of the IncQ plasmid pJAK13 (J. Kornacki, unpublished results). The region containing lacI^q and tacp was removed by digesting the plasmid with HpaI and HindIII, blunting the overhangs of the HindIII site using the Klenow fragment of DNA polymerase I (New England Biolabs), and ligating the blunt ends. Following transformation of E. coli TOP10 and plasmid isolation, we obtained pGHM485. The lacZ gene was amplified by PCR from pMLB1109 (a gift from M. Berman). KpnI and EcoRI sites were added to the 5′ and 3′ ends, respectively, during PCR. pGHM485 and the lacZ fragment were both digested with KpnI and EcoRI and ligated; then, E. coli TOP10 was transformed to create pJAKlac. This reporter plasmid contains a promoterless lacZ with an upstream multiple cloning site (Fig. 2A).

FIG. 2.

Reporter vector and analysis of the tad locus promoter. (A) pJAKlac was constructed as a promoterless transcriptional fusion vector using lacZ as a reporter. Sp^r indicates streptomycin and spectinomycin resistance. T_rrnB is the terminator for rRNA operon B. (B) Sequence of the CU1000 region upstream of the flp-1 translational start codon (italics). Predicted −35 and −10 sequences for tadp are underlined. The asterisk marks the putative transcriptional start site (+1). (C) β-Galactosidase assays for pJAKlac derivatives in CU1000N (bars a to d). Each strain harbors pJAKlac containing the following fragments in the multiple cloning site: bar a, empty vector; bar b, −80 to flp-1 translational start (+102); bar c, −80 to −1; bar d, −80 to +1. Percent transcription levels were compared to that from the −80-to-+102 construct, which was considered 100%. In each assay, samples were done in triplicate. The assays were done at least three times, with similar results each time. The error bars are for triplicate samples from one representative assay.

To construct some derivatives of pJAKlac, fragments were generated by PCR using primers that added an XbaI site and the tacp or tadp promoter to the 5′ ends and a BamHI site to the 3′ ends (Table 2). The tacp sequence was from the National Center for Biotechnology Information (13). Derivatives containing the promoter sequences were constructed with fragments generated by PCR using primers that added an XbaI site to the 5′ end and a BamHI site to the 3′ end of the promoter region (Table 2). The fragments and pJAKlac were digested sequentially with XbaI and BamHI and then gel purified using the Qiaex II gel extraction kit (Qiagen). They were ligated and used to transform E. coli TOP10. All constructs were confirmed by sequencing before mobilization into A. actinomycetemcomitans.

β-Galactosidase assays.

Assays on E. coli strains harboring reporter constructs were done as described previously (28). Briefly, E. coli overnight (∼20-h) cultures were inoculated into LB broth with appropriate antibiotics and 100 mM isopropyl β-d-thiogalactopyranoside (IPTG). These cultures were diluted 1:50 and grown to mid-exponential phase (optical density at 600 nm [OD₆₀₀], 0.4 to 0.7), and then 300 μl of culture was mixed with Z buffer (113 mM Na₂HPO₄, 40 mM NaH₂PO₄, 10 mM KCl, 1 mM MgCl₂, pH 7.0) (28), 50 μl chloroform, and 25 μl 0.1% sodium dodecyl sulfate to lyse the cells and extract β-galactosidase into the aqueous layer. Two hundred microliters of 4-mg/ml o-nitrophenyl β-d-galactopyranoside, a chromogenic substrate that is cleaved by β-galactosidase to form a yellow product, was then added to the samples. The aqueous layer was removed from the reaction tubes and read in a microtiter dish at A₄₂₀. β-Galactosidase expression was reported in Miller units (28).

The following changes were made to obtain accurate cell density measurements in A. actinomycetemcomitans. CU1000N- and CU1060N-derived strains were grown overnight (∼20 h) with 20 μg/ml spectinomycin to select the resident plasmid. IPTG was not necessary in these cultures, because A. actinomycetemcomitans does not contain a copy of the lacI repressor gene (http://www.oralgen.lanl.gov/oralgen/bacteria/aact/). OD₆₀₀ readings were taken for CU1060N strains, and cultures of similar density (OD₆₀₀ = 0.7 to 0.9) were used. CU1000N-derived strains were each scraped from the culture tube with a wooden dowel, harvested by centrifugation, and resuspended in 2 ml AAGM. One milliliter of this culture was harvested by centrifugation and resuspended in 100 μl of 1× Laemmli buffer (23). The cells were boiled for 10 min and diluted 1:100 for reading at A₂₈₀. This absorbance reading was used for CU1000N-derived strains instead of an OD₆₀₀ reading. Therefore, the units of these β-galactosidase assays were relative units instead of Miller units.

For assays to determine the effect of Rho on termination, bicyclomycin (BCM) (a gift of M. Gottesman) was added to E. coli TOP10 at a final concentration of 50 μg/ml for 40 min before lysis (as advised by R. Washburn and M. Gottesman) and to A. actinomycetemcomitans CU1060N at a concentration of 100 μg/ml 60 min before lysis. The higher concentration of BCM was used for CU1060N in the hope of accommodating the fact that this strain grows more slowly than E. coli. β-Galactosidase assays were then done as described above.

RNA extraction.

RNA was extracted from 16-hour overnight cultures of CU1000N and CU1060N using the Purescript RNA Purification System cell and tissue kit (Gentra). For CU1000N, 20 ml of AAGM was inoculated, and for CU1060N, 5 ml was used. Entire overnight cultures were harvested for RNA extraction from CU1000N, but only 1.5 ml for CU1060N. The Gentra kit yielded 1 to 5 μg/μl of high-quality RNA. To check the quality, samples were run on agarose-formaldehyde gels (6), along with an RNA size marker for reference (Invitrogen). Once extracted, RNA was stored at −80°C until it was used for Northern hybridization analysis.

Northern hybridization analysis.

Equivalent amounts of CU1000N or CU1060N RNA and an RNA marker (Invitrogen) were each denatured, separated by electrophoresis in an agarose-formaldehyde gel, and transferred overnight to a Hybond-N nitrocellulose membrane (Amersham Pharmacia Biotech). The membranes were baked for 1 hour at 80°C to immobilize the RNA and then hybridized with DNA probes corresponding to flp-1, flp-2, or tadV. Probes, which corresponded to nucleotides (from the translational start) 16 to 214 for flp-1, 5 to 230 for flp-2, and 1 to 246 for tadV were randomly labeled using Bio-14-CTP (Invitrogen), according to the manufacturer's instructions.

Sequence identification and secondary-structure predictions.

Sequences were identified and compared using National Center for Biotechnology Information BLAST functions (http://www.ncbi.nlm.nih.gov/) (1). Secondary-structure predictions were made using Sequencher software (Gene Codes Corporation) and MFOLD software (http://bioweb.pasteur.fr/seqanal/interfaces/mfold-simple.html).

RESULTS

A strong tad promoter.

The transcriptional start site for the tad locus was previously identified by rapid amplification of cDNA ends as 102 bp upstream of the flp-1 translational start site (15). A promoter was inferred from both this result and the nucleotide sequence of the region (15) (Fig. 2B). The −10 region of this promoter (tadp) agrees with the consensus σ⁷⁰ sequence, and the −35 region differs by only 1 base. tadp was shown to be sufficient for transcriptional initiation in E. coli (15). We determined the sequence of the tadp region in the CU1060N nonadherent, smooth strain. Its tadp has a mutation in the −10 region in a base that is highly conserved in σ⁷⁰ promoters (reference 24 and data not shown). The mutant promoter was predicted to initiate transcription poorly, if at all, based on known promoter sequences (24). CU1060N is nonadherent and nonpiliated, and no Tad proteins are visible by Western blot analysis (B. Perez, S. Clock, and D. Figurski, unpublished results, and data not shown). Taken together, these results suggest that wild-type tadp has the properties expected for expression of the tad locus in A. actinomycetemcomitans.

We wished to know if tadp is functional in A. actinomycetemcomitans. Since there were no reports of a transcriptional reporter plasmid for this organism, we constructed pJAKlac, a plasmid that replicates in A. actinomycetemcomitans and harbors a promoterless lacZ gene downstream of a multiple cloning site (Fig. 2A). The lacZ gene product could be measured by β-galactosidase assays (see Materials and Methods).

By cloning different parts of the tadp region, we determined that tadp is sufficient for transcription of lacZ in A. actinomycetemcomitans. No intrinsic β-galactosidase activity is present in A. actinomycetemcomitans (data not shown), and no expression was observed from the pJAKlac vector (Fig. 2C, bar a). A transcriptional fusion with a small fragment containing the intact tadp region (−80 to +102) upstream of lacZ exhibited a high level of β-galactosidase (Fig. 2C, bar b). Expression of lacZ was abolished when the tadp region (−80 to −1) did not contain the transcriptional start site (Fig. 2C, bar c), but it was seen when the construct contained it (−80 to +1) (Fig. 2C bar d). These data led us to believe that tadp is functional in A. actinomycetemcomitans.

We also wanted to examine the strength of tadp relative to a known promoter. Since tadp was thought to be a strong promoter because of its σ⁷⁰-like consensus regions, minimal tadp and tacp promoters were each cloned into pJAKlac (data not shown). We used the minimal tadp promoter region that extended from the −35 site through the transcription start site. It might have been reduced in strength relative to the one that begins at −80, but we found it to be sufficient for transcription initiation. No lacI gene, which encodes the lactose repressor, exists in A. actinomycetemcomitans (http://www.oralgen.lanl.gov/oralgen/bacteria/aact/), so tacp is constitutive in the organism. When β-galactosidase activities were measured, tadp was at least as strong as unrepressed tacp, if not slightly stronger (data not shown). Because tacp is known to be a strong synthetic promoter (14), these results indicate that tadp is also a strong promoter. We also compared tadp and tacp in E. coli, where the synthetic tacp is known to be strong. Here, too, tadp was as strong as tacp (data not shown).

Identification of tad transcripts.

Past Northern blots of A. actinomycetemcomitans were mostly from nonadherent (smooth) strains, and those from adherent (rough) strains did not show distinguishable transcripts (15, 31). We used another approach to extract intact RNA from adherent (rough) strains (see Materials and Methods). This approach allowed us to identify pilin (flp-1) transcripts by Northern blot analysis.

If the tad locus is expressed as a single polycistronic mRNA, there should be a transcript of approximately 12 kb. We did not detect a large transcript by Northern blot analysis, possibly due to its low abundance or instability or to the possibility that tad may not be expressed as a polycistronic unit. We did identify two lower-molecular-weight transcripts. An abundant transcript of approximately 0.4 kb hybridized only to a flp-1-derived probe (Fig. 3). Another transcript of approximately 1.1 kb was less abundant and hybridized to probes derived from the flp-1, flp-2, and tadV genes (Fig. 3). Because these genes are at the beginning of the tad locus, we tested genes further downstream. tadA and tadG probes did not hybridize to these small transcripts (data not shown). The small transcripts were not present in CU1060N, indicating that they originated from tadp (data not shown). We considered that the small transcripts resulted from transcriptional termination or from posttranscriptional mRNA processing of a larger transcript. In the former model, if the observed transcripts were initiated at tadp, the transcript length indicated termination after flp-1 (0.4 kb) and after tadV (1.1 kb).

FIG. 3.

Northern hybridization analysis. RNA extraction was done at 20 h of growth. Probes for flp-1, flp-2, and tadV were used, as described in Materials and Methods.

Secondary-structure analysis.

We analyzed the nucleotide sequences of the intergenic regions after flp-1 and tadV (R1 and R2, respectively), where the transcripts may have terminated, using the sequence of CU1000N (accession no. AY157714). In the intergenic region downstream of flp-1, we identified a putative stem-loop structure similar to a classical intrinsic terminator, including a downstream run of T and A nucleotides in the DNA (Fig. 4). The calculated ΔG of this putative intrinsic terminator was −13.98 kcal, which is in the same range as published ΔG values for well-known strong intrinsic terminators (49). We designated this putative terminator T1. The putative stem-loop is present in other strains of A. actinomycetemcomitans, including D7S and DF2200 (serotype a), HK1651 (serotype b), and IDH781 (serotype d) (Fig. 4). We also noted that whenever a nucleotide in the putative stem was different, its partner was also different, so that base pairing was possible. This property is common among intrinsic terminators (49).

FIG. 4.

Predicted secondary structures. Predicted stem-loops, which are conserved in all of the different strains of A. actinomycetemcomitans examined (shown above the structures), are shown for the R1 and R3 regions. ΔG values for strain CU1000 were predicted to be −13.98 kcal (for the putative stem-loop in R1) and −15.5 kcal (for the putative stem-loop in R3).

We identified a weak (ΔG = −0.79 kcal) putative stem-loop in the intergenic region downstream of tadV that was conserved among other strains of A. actinomycetemcomitans (data not shown). This putative stem-loop is unlikely to be an intrinsic terminator. However, it is possible that factor-dependent termination occurs in the region or that the RNA is shortened by a processing event. The terminator in the R2 region has been designated T2.

tadG is the gene at the downstream end of the tad locus, and it is likely that transcription termination occurs shortly after it to prevent transcriptional read-through into adjacent genes. We found a possible intrinsic terminator sequence in the region downstream of tadG (R3). Like R1, R3 has a stem-loop with a low calculated ΔG (−15.5 kcal) that is conserved in other strains of A. actinomycetemcomitans (Fig. 4). We designated this putative terminator T3.

Termination efficiencies of T1, T2, and T3.

To determine if R1, R2, and R3 led to transcriptional termination, they and their derivatives were cloned into a multiple cloning site between lacZ and tacp in our reporter plasmid (Fig. 5A). β-Galactosidase assays were done on strains carrying the reporter plasmids with or without the putative terminators.

FIG. 5.

Analysis of termination within the tad locus. (A) Fragments placed into the pJAKlac reporter vector upstream of lacZ to make constructs, as described in Materials and Methods. The small opposing arrows represent inverted repeats, which make up the stems of the putative stem-loops identified in regions R1, R2, and R3. R1⁺, R2⁺, and R3⁺ constructs contain these stem-loops, while R1⁻, R2⁻, and R3⁻ do not. R2′ contains part of the intergenic region, but not the stem-loop. Addition of those fragments created transcriptional fusions for use in β-galactosidase assays, which were done in E. coli (B, D, and F) and in A. actinomycetemcomitans CU1000N (C, E, and G). In all graphs, percent transcription is normalized to transcription from the “−” construct, which lacks the terminator. The R2⁺ region shows increased expression in A. actinomycetemcomitans (see the text and Table 3). Note that the ordinate in panel E is different to allow visualization of the increased transcription. In each assay, samples were done in triplicate. The assays were done at least three times, with similar results each time. The error bars are for triplicate samples from one representative assay. The asterisks marks P values of <0.05 as calculated by Student's t test.

T1 is predicted to be a strong intrinsic terminator, based on a putative stem-loop with a low calculated ΔG value (Fig. 3 and 4A). β-Galactosidase assays of the R1 constructs were done in both E. coli and A. actinomycetemcomitans, because T1 should function in both. We found that termination occurred with 90% to 99% efficiency in both E. coli and A. actinomycetemcomitans relative to transcription from the R1⁻ construct (Fig. 5B and C and Table 3).

TABLE 3.

Termination efficiencies^a

Region (terminator)	Efficiency (%)
	E. coli	A. actinomycetemcomitans
R1+ (T1)	98	99
R2′ (T2)	50	36
R2+ (T2)	50	NCb
R3+ (T3)	69	86

^aThe termination efficiencies were calculated from β-galactosidase assays in which strains with terminator-plus reporter plasmids were compared to those with terminator-minus reporter plasmids.

^bNC, not calculated. β-Galactosidase is more highly expressed in A. actinomycetemcomitans when the R2⁺ region is compared to the R2⁻ region (see the text and Fig. 5E).

To examine T2, we created three constructs. R2⁺ contained the entire intergenic region, including the weak putative stem-loop, while R2⁻ did not contain any of the intergenic region. R2′ was a truncation that included part of the intergenic region between tadV and rcpB but did not contain the putative stem-loop. β-Galactosidase assays were done in E. coli and A. actinomycetemcomitans. Our results showed that R2′ terminated 36% to 50% of the expression of β-galactosidase in both E. coli and A. actinomycetemcomitans compared to expression from R2⁻ (Fig. 5D and E and Table 3), indicating that the R2′ fragment contained T2, the region necessary to reduce expression, even though it lacked a putative stem-loop. In E. coli, the R2⁺ construct also showed 50% termination, while the same construct showed increased expression in A. actinomycetemcomitans (Fig. 5F and G). One possible explanation for the increase is the existence of a promoter in the second part of R2. Such a promoter would be predicted to initiate transcription in A. actinomycetemcomitans, but not in E. coli. We identified no obvious sequences related to the consensus sequences for the σ⁷⁰, σ³², or σ²⁴ promoter (data not shown). These σ factors are known to be present in the genome of A. actinomycetemcomitans (http://www.oralgen.lanl.gov/oralgen/bacteria/aact/). Alternatively, increased expression could also be explained by an antitermination activity that occurs in A. actinomycetemcomitans but not in E. coli. Ongoing studies in our laboratory will seek to distinguish between these possibilities.

Termination due to T3 is predicted to occur at the end of the tad locus to prevent read-through of RNA polymerase into neighboring genes. β-Galactosidase assays showed that R3 terminates strongly, as expected. There was a slight difference in termination by R3 in E. coli and A. actinomycetemcomitans (69% versus 86%, respectively) (Fig. 5F and G and Table 3). This difference was reproducible and may be indicative of less termination at this site in E. coli.

Rho and T2.

In E. coli and other bacteria, Rho is a protein that binds nascent mRNAs and terminates transcription through an ATP-dependent mechanism (36). No specific Rho recognition sequence has been found, but binding frequently occurs within cytosine-rich stretches in transcripts (36). There is a homolog of rho in A. actinomycetemcomitans (http://www.oralgen.lanl.gov/oralgen/bacteria/aact/). The protein is predicted to be 88% identical and 95% similar to Rho in E. coli (data not shown). Because the R2′ termination region does not contain any significant secondary structure, we tested the possibility that Rho might play a role in T2 termination.

We used the antibiotic BCM, which inhibits the ATPase domain of Rho, to disable its transcription termination functions (42). We used A. actinomycetemcomitans CU1060N in this experiment because it is a nonadherent derivative of CU1000 that is easier to grow and assay than is CU1000 and it behaved the same in other termination experiments (data not shown). BCM inhibits the growth of A. actinomycetemcomitans CU1060N, as demonstrated by a very low MIC (<3 μg/ml) relative to the reported MIC of BCM in E. coli of 25 μg/ml (39). If termination by T2 is Rho dependent, R2′-containing reporter constructs should not terminate upon addition of BCM and should have the same expression level as R2⁻. In contrast, termination by T1 and T3 should be unchanged, as they are both predicted to terminate by a Rho-independent mechanism.

Our results showed that upon addition of BCM, termination in the R2′ region was inhibited completely in both A. actinomycetemcomitans and E. coli (Fig. 6 and data not shown). In contrast, there were no significant differences in the termination efficiencies of R1 and R3 (data not shown). These results indicate that T2 termination is dependent on Rho, while T1 and T3, predicted to be intrinsic terminators, are not. These results further support the hypothesis that the R2′ region contains information that allows Rho-based transcriptional termination.

FIG. 6.

BCM inhibition of T2 termination. β-Galactosidase assays were done in CU1060N strains containing plasmids with either the CU1000N R2′ region, which terminates, or the CU1000N R2⁻ region, which does not terminate, as described in Results and Materials and Methods. BCM was added after an overnight incubation, 60 min before lysis. One hundred percent termination by T2 was calculated from R2′ in cells grown without BCM. Thirty-six percent reduction of transcription was equal to 100% termination for T2. In each assay, samples were done in triplicate. The assays were done at least three times, with similar results each time. The error bars are for triplicate samples from one representative assay.

DISCUSSION

We have shown that a promoter (tadp) at the beginning of the tad locus has strong activity in A. actinomycetemcomitans. Our data also indicate that differential expression of some genes of the tad locus involves a transcriptional-termination cascade. Using Northern hybridization analysis, sequence analysis, and β-galactosidase assays, we identified three regions at which termination occurs. In contrast to T1 and T3, which are likely to be intrinsic terminators, T2 lacks a low-energy stem-loop structure, and termination is inhibited by BCM. T2 seems to act through a Rho-dependent mechanism.

The −35 and −10 elements of tadp are similar to the σ⁷⁰ consensus sequence for E. coli, indicating that the promoter is strong. This possibility was confirmed by comparing it to unrepressed tacp in A. actinomycetemcomitans and E. coli. Its location before flp-1 indicates that it could be responsible for transcription of the entire tad locus. Our nonadherent variant, CU1060N, has a mutation in the −10 sequence of tadp. Other work has shown that mutations in the −10 sequence of tadp correlate with loss of adherence (48). Consistent with these data is the lack of detectable transcripts in our Northern blot analysis and the lack of detectable proteins by Western blot analysis in CU1060N, implying that the tad genes are expressed poorly or not at all in this strain (Perez et al., unpublished, and data not shown).

Two models could explain the tad transcripts seen in our Northern hybridization analysis. The first model holds that processing of mRNA is followed by stabilization. It is possible that the stem-loop structures stabilize transcripts by protecting against degradation by 3′ exoribonucleases (3, 44). Indeed, several pilus systems rely on processing in conjunction with differential stability of the mRNA (2, 3, 5, 19, 29, 44). For example, in uropathogenic E. coli, S fimbriae are encoded by the sfa locus of nine genes (3). The fimbrial-subunit gene, sfaA, is the third gene in the locus, but it is expressed alone as the major transcript (3, 29). This expression pattern is achieved by mRNA processing and degradation. In addition, partial termination by an intrinsic terminator downstream of the sfaA gene has also been hypothesized (3).

Our data support a second model in which transcripts initiated at tadp are terminated differentially. We showed that the intergenic region of CU1000 between flp-1 and flp-2 was sufficient to terminate 99% of transcription, roughly accounting for the differences in transcript abundance seen in the Northern hybridization analysis.

Termination within polycistronic transcriptional units in bacteria can regulate the stoichiometry of proteins. In the macromolecular synthesis operon, consisting of rpsU-dnaG-rpoD in E. coli and other gram-negative bacteria and some gram-positive bacteria, an intrinsic terminator in the intergenic region between rpsU and dnaG partially regulates the differential expression of the upstream gene relative to downstream genes (25). Transcriptional termination has also been shown to be important in several amino acid biosynthetic operons (50).

The predicted 12-kb transcript corresponding to the entire tad locus has not been detected. It is possible that the entire locus may not be transcribed as a single unit. Although transcription of the intergenic regions of the tad locus has been shown from the beginning of the locus through tadD (15), transcription of downstream genes has not been reported. Other explanations for the inability to detect a 12-kb transcript are that there may be rapid processing events so that the transcript is never wholly present, the transcript may be unstable, or the transcript may be transcribed at a level below the limit of detection.

T2 termination seems to operate through a mechanism different from that of T1 and T3. BCM, a Rho inhibitor (42), inhibits T2 termination in both E. coli and A. actinomycetemcomitans, but transcription termination at T1 and T3 is unchanged. Since no significant secondary structure or specific sequence is predicted for T2, termination at T2 likely occurs through a factor-dependent mechanism, probably through Rho.

The increased expression of R2⁺ over R2⁻ in A. actinomycetemcomitans may be due to an internal promoter within the region that is recognized only in A. actinomycetemcomitans. Alternatively, it could be due to an antitermination factor that is active in A. actinomycetemcomitans and not in E. coli. NusG, a Rho-dependent antitermination factor in E. coli that is also encoded in the genome of A. actinomycetemcomitans (http://www.oralgen.lanl.gov/oralgen/bacteria/aact/), may be responsible for this activity in A. actinomycetemcomitans as well. To explain the different responses of E. coli and A. actinomycetemcomitans, it is not difficult to imagine that the NusG proteins of these bacteria have diverged to recognize different sequences.

We have demonstrated that differential expression of genes in the tad locus of A. actinomycetemcomitans is achieved, at least partially, by a transcriptional-termination cascade. Our results show three terminators, although additional terminators may exist. Northern blot analysis has clearly revealed two transcripts, but other results indicate there are at least three transcripts (15) (Fig. 7). The estimated stoichiometry of the Tad proteins correlates with the number of transcripts predicted by the model in Fig. 7. Assuming that tad genes are expressed from a polycistronic message, for every 12-kb transcript, there would be approximately 1.6 flp-1 plus flp-2 plus tadV (1.1-kb) transcripts and 156 flp-1 (0.4-kb) transcripts. These ratios roughly correlate with the differences seen in Northern blot analysis (Fig. 3). In agreement with this model, we found that, of all the expressed tad genes, only flp-1 needed to be induced from tacp for complementation (20, 35). Slightly more TadV would help process the Flp prepilin subunits, while proteins that compose the structural components of the Tad machinery might be needed in small amounts (43). These data, taken together, indicate that the transcription of the tad locus is regulated, at least in part, by a transcriptional-termination cascade.

FIG. 7.

Model for transcriptional regulation within the tad locus. Sites of intrinsic termination (T1 and T3) are shown by lollipops, and the region for Rho-dependent termination (T2) is shown by a bracket. The putative 12-kb transcript is shown as a dashed line, since it has not been visualized. All three transcripts are assumed to initiate from tadp. Landmark genes of the tad locus are represented by the gray arrows.