Characterization of the glnK-amtB Operon of Azotobacter vinelandii

Abstract

To determine whether in Azotobacter vinelandii the P_II protein influences the regulation of nif gene expression in response to fluxes in the ammonium supply, the gene encoding P_II was isolated and characterized. Its deduced translation product was highly similar to P_II proteins from other organisms, with the greatest degree of relatedness being exhibited to the Escherichia coli glnK gene product. A gene designated amtB was found downstream of and was cotranscribed with glnK as in E. coli. The AmtB protein is similar to functionally characterized ammonium transport proteins from a few other eukaryotes and one other prokaryote. glnK and amtB comprise an operon. Attempts to isolate a stable glnK mutant strain were unsuccessful, suggesting that glnK, like glnA, is an essential gene in A. vinelandii. amtB mutants were isolated, and although growth on limiting amounts of ammonium was similar in the mutant and wild-type strains, the mutants were unable to transport [¹⁴C]methylammonium.

Nitrogen fixation genes (nif) are highly conserved among all nitrogen-fixing bacteria, and in all diazotrophic species of the class Proteobacteria examined, the transcriptional activator NifA is required for expression of other nif genes (33). In some diazotrophs of the α and β subgroups of the class Proteobacteria, such as Azospirillum brasilense, Rhodobacter capsulatus, and Herbaspirillum seropedicae, the NifA proteins are inactive in cells grown with high levels of ammonium (1, 4, 31). The mechanism(s) for this is unknown but may involve a P_II protein, known from studies of enteric bacteria to be a major component of a nitrogen-sensing and signal transduction cascade. In diazotrophs of the proteobacterial γ subgroup, including Azotobacter vinelandii and Klebsiella pneumoniae, a nifL gene lies upstream of nifA (3, 5, 18, 26). NifL interacts with and prevents the activity of NifA in cells exposed to oxygen or excess fixed N. Inactivation of NifA prevents it from stimulating transcription from promoters adjacent to the several other nif genes and operons, leading to a failure to produce nitrogenase enzyme. K. pneumoniae NifA deleted of its N-terminal domain is more sensitive to inactivation by NifL, suggesting that a function of this domain is to modulate the response of NifA to NifL (12). NifL of A. vinelandii was recently shown to be a redox-sensitive flavoprotein with flavin adenine dinucleotide as the prosthetic group, which when reduced, has no effect on in vitro NifA-dependent open-complex formation at the nifH promoter (17). When oxidized, NifL prevents open-complex formation. How NifL responds to fixed-N status by becoming inhibitory to NifA in is not known. It is also possible that the susceptibility of NifA to NifL inactivation increases as the fixed-N supply increases.

Transduction of the environmental signal of fixed-nitrogen status has been best described in enteric organisms (for a review, see reference 27). Under low-ammonium-concentration conditions, the product of the glnB gene, the P_IIB protein, a homotrimer consisting of identical 12.4-kDa subunits (9), is uridylylated at a Tyr residue by the uridylytransferase activity of the glnD gene product (32). The uridylylation state of P_II determines whether the transcriptional activator NtrC is phosphorylated or dephosphorylated (i.e., active or inactive, respectively) (2, 29) and the extent to which glutamine synthetase (GS) is adenylylated or deadenylylated (i.e., inactive or active, respectively) (20). The recent finding of second P_II-encoding genes, i.e., glnK (present in addition to glnB in Escherichia coli [38]) and glnZ (present in addition to glnB in Azospirillum brasilense [10]), complicates the picture, especially with respect to what occurs under low-fixed-N conditions.

In the nonenteric bacterium A. vinelandii, the glnD gene, originally named nfrX, was identified by Tn5 mutagenesis (34) and subsequent DNA sequence analysis (8). While glnD::Tn5 mutants are Nif⁻, glnD::Tn5 nifL::KIXX double mutants are Nif⁺. Two models developed to explain this result are as follows: (i) GlnD is required for conversion of active NifL (which inhibits NifA) to its inactive form, and (ii) GlnD is necessary for conversion of NifA to a conformation that cannot be inactivated by NifL. The question of whether GlnD influences NifL (or NifA) via a P_II protein was the basis for this study.

Isolation and sequencing of the A. vinelandii glnK and amtB genes.

The glnB gene of E. coli on pAH5 (19) hybridized to a 3.2-kb EcoRI fragment of A. vinelandii genomic DNA on Southern analysis (data not shown). This fragment was cloned in pACYC184 (6), giving pND183, and subsequently into pSVB30 (25), giving pDM508. The glnB-hybridizing region was further delineated by hybridization. The nucleotide sequence of the 2.3-kb PstI-EcoRI fragment of pPR101 (Fig. 1) was determined on both strands by using exonuclease III- and nuclease S1-generated deletion derivatives. Analysis of the 2,278-bp sequence revealed two potential open reading frames (ORFs), one between nucleotide positions 477 (ATG) and 813(TGA) and one between nucleotides 849 (ATG) and 2,157 (TGA). These ORFs potentially encode proteins with molecular weights of 12,240 and 46,390 Da, respectively. The product of the first ORF showed a high degree of overall amino acid sequence similarity to several P_II proteins and includes the conserved site of uridylylation at Tyr51. The most similar of the other proteins is the second P_II protein in E. coli, designated GlnK; the two proteins are 75% identical and 84% similar with respect to amino acid sequence (data not shown). E. coli GlnB is less similar, with 66% identical amino acids. High degrees of similarity between the glnB gene products of other members of the class Proteobacteria and the product of the first A. vinelandii ORF described here also are evident, with identities ranging from approximately 70 to 74%. Because of the high degree of similarity of this ORF product to the E. coli glnK gene product and the proximity of this P_II protein-encoding gene to a putative ammonium transporter in both organisms (see below), the A. vinelandii gene has been designated glnK.

FIG. 1.

The 2.3-kb glnK-amtB region of A. vinelandii. The locations of KIXX and lacZ interposon constructs and their directions of insertion are shown by the hatched rectangles and arrows below the line, respectively. The resulting plasmids are named along with (where appropriate) the stable mutant strain of A. vinelandii in which the glnK::KIXX or glnK::lacZ region replaced the wild-type glnK gene. The fragments used as probes to identify RNA fragments on the Northern blots (see Fig. 2) are shown as thin lines. Abbreviations for restriction sites: B, BamHI; E, EcoRI; H, HincII; Sn, StuI; SnaBI; P, PstI.

The second product of the second A. vinelandii ORF exhibits the highest degree of overall similarity to the NrgA protein of Bacillus subtilis (39), with 41% identical and 56% similar amino acids. While no specific function for NrgA has been determined in this organism, the nrgA gene lies adjacent to and upstream of a glnB-like gene named nrgB, in contrast to the arrangement identified here for A. vinelandii, in which glnK is upstream of the nrgA-like gene. During the course of this work, five other genes encoding NrgA-like proteins were described for two eukaryotes and two other prokaryotes, and their products are also similar to that of the A. vinelandii ORF described here (with 25 to 38% identity). These proteins have a significant hydrophobic core, with 10 to 12 transmembrane-spanning domains flanked by hydrophilic regions, and include products of the mep1, mep2, and mep3 genes from Saccharomyces cerevisiae (22, 23), the amt1 gene of Arabidopsis thaliana (30), the amt gene of Corynebacterium glutamicum, and the amtB gene of E. coli (35, 38). Of these organisms, the only one showing a definite growth phenotype was a mep1 mep2 double mutant of S. cerevisiae which is unable to grow on media with low levels of exogenous ammonium (13). While there are no mutants of the amt1 gene of Arabidopsis thaliana available for comparison, the cloned gene from this organism was able to complement the S. cerevisiae mutant for growth on media with low levels of ammonium. Mutations in the amt gene of C. glutamicum transported 20-fold less [¹⁴C]methylammonium than did the amt⁺ parent strain. Thus, it seemed possible that the nrgA-like gene of A. vinelandii located downstream of glnK might encode a membrane-spanning protein involved in ammonium transport (see below). This gene has been named amtB to distinguish it from amtA, an unrelated gene initially identified as being involved in ammonium transport in E. coli (14) but now thought to encode CysQ (28).

glnK and amtB gene expression.

Potential ribosome binding sites for glnK and amtB were located within 10 bp upstream of the ATG initiation codons. The small intergenic region between glnK and amtB as well as the lack of any obvious transcriptional terminator- or promoter-like structures suggested the probable cotranscription of glnK and amtB. Sequences similar to the −10- and −35-like regions of other prokaryotic genes recognized by ς⁷⁰ are present from 36 to 66 bp upstream of the ATG of glnK. Whether these sequences are significant for transcription of the putative glnK-amtB operon was not investigated. About 75 bp downstream of amtB there begins a 27-bp region with features characteristic of a factor-independent transcription terminator, including a 6-base inverted-repeat motif separated by 3 bp and followed by a tract of six T’s.

Northern blotting experiments were carried out to determine whether fixed N regulates expression of the glnK and amtB genes and whether these two genes are cotranscribed. RNA prepared from either ammonium- or N₂-grown A. vinelandii cultures hybridized extremely weakly but reproducibly to three different [³²P]dCTP-labeled probes: a 360-bp HincII fragment (glnK only), a 1.2-kb StuI fragment (glnK and amtB), and a 0.9-kb StuI fragment (amtB only) (Fig. 1). In all cases, the hybridizing RNA band corresponded to a transcript of approximately 1.7 kb, the size expected if the glnK and amtB genes are cotranscribed (Fig. 2). The intensity of labeling of this band was slightly less for N₂-grown cultures than for NH₄⁺-grown cultures (Fig. 2). To determine whether the amount of glnK-amtB mRNA present was unusually small compared to that of other transcripts, the same Northern blots were hybridized with a glnA gene probe. The intensity of the hybridizing band observed at 1.5 kb, the size of the glnA transcript previously observed (36) after autoradiography, was about 100-fold greater than that of the glnK-amtB band (data not shown). glnK-amtB expression was also measured as the amount of β-galactosidase produced in an amtB-lacZ fusion strain, MV566, constructed by transformation of UW136 with pPR203 (Fig. 1) grown under different conditions. The levels of expression were consistently very low, about 75 Miller units of activity, and were not significantly different in ammonium-, urea-, or N₂-grown cultures.

FIG. 2.

Northern blot analysis of the glnK-amtB operon. mRNA from ammonium-grown (lanes a, c, and e) or N₂-grown (lanes b, d, and f) cultures was prepared by using the Qiagen (Chatsworth, Calif.) RNeasey Kit (catalog no. 74904). Samples containing approximately equal amounts of total RNA were separated on 0.8% formaldehyde–agarose gels by electrophoresis; this was followed by capillary blotting of the RNA onto nylon membranes. Blots were hybridized to three ³²P-labeled probes from the glnK-amtB region: a 360-bp HincII fragment (glnK only), a 1.2-kb StuI fragment (glnK plus amtB), and a 0.9-kb StuI fragment (amtB only). L, DNA fragment standards.

These experiments show that the glnK and amtB genes are cotranscribed in an operon and that their levels of expression are very low and not significantly influenced by the fixed-N supply. In contrast, the glnK-amtB operon of E. coli is transcribed from an NtrC-activated promoter and is therefore not expressed in cells grown with high levels of ammonium (38). The glnB genes of several other organisms, including Rhizobium etli, Rhodobacter capsulatus, and Rhodospirillum rubrum, are also dependent on NtrC for expression (14a, 21a, 31a). So far, A. vinelandii glnK, E. coli glnB, and Azospirillum brasilense glnZ are the only known non-fixed-N-regulated P_II-encoding genes among the members of the class Proteobacteria (10, 37a). The absence of both ς⁵⁴ and NtrC recognition motifs in the A. vinelandii glnK promoter region is consistent with the results obtained from the expression experiments.

Whether the glnK gene in A. vinelandii reported here represents the only P_II-encoding gene in this organism is uncertain. There is some evidence that there is not a second functional P_II protein encoded by another gene; only a single hybridizing band was observed after hybridization of either the E. coli glnB gene or the A. vinelandii glnK gene, characterized here, to genomic digests generated with three different restriction enzymes. In an experiment involving the cloning of P_II-encoding genes by ligation of PCR products generated from oligonucleotide primers based on conserved amino acid sequences in both GlnK and GlnB P_II proteins, each of the 10 products cloned had a DNA sequence identical to that of the glnK gene described here (24). Also, as discussed below, glnB mutants could not be isolated under a variety of growth conditions, including those under which another glnB-like gene might be expected to be expressed.

Attempts to construct glnK mutants of A. vinelandii.

The kanamycin resistance-encoding KIXX cassette was inserted at or between endonuclease restriction sites in pPR101 (Fig. 1). A. vinelandii UW136 was transformed with the three resulting glnK::KIXX plasmids, pDM513, pDM514, and pPR401. The kanamycin-resistant, ampicillin-sensitive transformants, which presumably carried the desired gene replacements because of the occurrence of a double-crossover event, were serially subcultured three times on selective medium containing kanamycin. To verify that replacement of the chromosomal wild-type copies of glnK had occurred, genomic DNA was isolated and analyzed in Southern hybridization experiments using pDM508 as a probe. However, in all transformants examined, both wild-type and mutated copies of the glnK::KIXX genes were detected even after prolonged growth of up to 10 subcultures under selective conditions (i.e., with kanamycin) (data not shown). Also, colonies isolated after 10 cycles of selection followed by 1 cycle of growth on medium without kanamycin had all become sensitive to the antibiotic. The same pattern of behavior was observed if transformed cells were plated on a medium with or without ammonium or with poor or excellent carbon sources or were incubated under aerobic or microaerobic growth conditions. Therefore, these mutants showed aberrant behavior similar to that observed when attempts were made to construct glnA insertion mutants of A. vinelandii (36). From the results of this previous work it was concluded that glnA null mutations are lethal because GS is the only ammonium assimilation-associated enzyme present and glutamine cannot be transported in A. vinelandii. It thus appears that null mutations in glnK also cannot be tolerated and are lethal. Similar results were obtained in attempts to construct glnD::KIXX null mutants (7). (While the original glnD [orignally named nfrX] mutants were viable and Nif⁻, the sites of Tn5 insertion were at the far 3′ end of the gene, leaving doubt as to whether they represented null mutants [see references 8 and 36]). Since P_II-UMP is required in enteric bacteria for rapid deadenylylation of GS, one possible reason for severe growth impairment in the A. vinelandii glnK (and also glnD) mutants is that GS is insufficiently active (i.e., remains adenylylated even under low-fixed-N conditions). To test this possibility, MV75, a strain in which GS cannot be adenylylated carrying a glnA gene in which Tyr407 has been mutated to Phe, (15), was transformed with the glnK::KIXX plasmids pDM513 and pDM514 individually. While the altered GS state did allow the glnD mutations to become stable and complete gene replacement occurred (7), the glnK::KIXX MV75 transformants behaved like the wild-type transformants in that wild-type chromosomes were always present after many subcultures on antibiotic-containing medium and resistance was quickly lost if the colonies were plated on antibiotic-free medium. Therefore, the lethality or impairment of growth of A. vinelandii caused by the introduction of glnK mutations is possibly due to an effect not only on GS activity but also on some other vital cellular function. As indicated above, there is unlikely to be a second P_II-encoding gene in A. vinelandii, as there is in E. coli, Azospirillum brasilense, and certain other bacteria. It is of relevance here that glnB glnZ double mutants of Azospirillum brasilense are severely growth impaired (11), as are E. coli glnB glnK mutants (16a), and that glnB mutations also appear to be lethal both in Synecocchus spp. and in Rhodospirillum rubrum since standard genetic techniques failed to result in replacement of the wild-type gene with a mutated copy (16, 21a).

amtB mutants are unable to transport [¹⁴C]methylammonium.

To construct amtB interposon mutations, pDM508 was partially digested with StuI and ligated to the KIXX cassette that had been blunt-ended with SmaI, giving pPR201 and pPR202 (Fig. 1). Wild-type strain UW136 and MV101 were transformed with both plasmids, and this was followed by selection on kanamycin-containing medium. Complete replacement of the amtB gene was confirmed by Southern analysis for both KIXX insertion transformants of each strain (data not shown); all were stably kanamycin resistant even after several subcultures in nonselective medium. Thus, unlike glnK, the amtB gene has no function that is of vital importance to A. vinelandii.

It was hypothesized that if the amtB gene encodes an ammonium transport protein, then the amtB::KIXX mutant strains may be less able to grow on limiting ammonium concentrations than the parental wild-type strain. The amtB mutant strains MV560 and MV561 were able to fix nitrogen and grew only slightly less well than UW136 on N-free BS agar medium. Therefore, the Nif⁻ nifH::KSS (lacZ) amtB::KIXX mutant strains MV562 and MV563 (Fig. 1) were used in these experiments since their growth is dependent on a supply of fixed nitrogen. Mutants and parent strain MV101 were grown in BS medium plus urea (5 mM) and then diluted and plated by pipetting 20-μl suspension aliquots, each containing approximately 200 cells, onto BS medium containing ammonium acetate at concentrations ranging from 0.05 to 10 mM. There was no difference in the results obtained with the two strains; in both cases, the rate of colony formation and the size of the colonies were the same at all concentrations and the colony size decreased with decreasing ammonium concentrations, becoming barely discernible at an ammonium concentration of about 100 μM.

Another test for ammonium transport is the uptake of the radiolabeled ammonium analog [¹⁴C]methylammonium. In these experiments, which required aeration of small sample volumes to detect transport in the wild-type strain, UW136 (amtB⁺) accumulated 10.6 nmol of methylammonium after 35 min (Fig. 3). In contrast, the amtB::KIXX mutant strain, MV560, failed to transport methylammonium (taking up less than 5% of the amount transported by the wild-type strain). The amtB⁺ strain UW136 failed to transport [¹⁴C]methylammonium if 15 mM ammonium was added at the beginning of the experiment, as expected since ammonium was shown to be a competitive inhibitor of methylammonium transport (21). In another control experiment, in which two or three samples were washed on the filters, washing never removed more than 20% of the bound radioactivity, indicating that the retained ¹⁴C was due to true uptake and not to adventitious binding of [¹⁴C]methylammonium. A clear phenotype of the amtB::KIXX mutants is therefore evident: they cannot transport methylammonium. While a similar phenotype was reported for amt mutants of C. glutamicum (35) and Azospirillum brasilense (37), the function of other prokaryotic AmtB proteins with respect to ammonium or methylammonium uptake has not been reported, nor (with the exception of an S. cerevisiae strain mutated in both mep genes) has an amtB mutant been shown to be deficient in ammonium transport per se. While it is widely assumed that methylammonium is transported as an analog of ammonium, this may not always be the case, and the determination of the true physiological function of amtB in prokaryotes must at least await identification of a specific physiological phenotype associated with amtB mutants. It is also likely that at least some prokaryotes contain two more ammonium transporters, each with a certain affinity for substrate, as in S. cerevisiae.

FIG. 3.

[¹⁴C]Methylammonium uptake by wild-type (▪) and amtB mutant (▴) strains. Overnight cultures in Burke’s N-free sucrose medium were diluted 50-fold and grown to an optical density at 600 nm of 0.5 to 0.7. The details of the uptake experiments were described by Jayakumar and Barnes (21).

Nucleotide sequence accession number.

The nucleotide sequence for the 2.3-kb region of pPR101 carrying the glnK and amtB genes of A. vinelandii has been deposited in the GENEMBL database under accession no. U91902.