The Cryptic Adenine Deaminase Gene of Escherichia coli

In Escherichia coli there are two pathways for conversion of adenine into guanine nucleotides, both involving the intermediary formation of IMP. The major pathway involves conversion of adenine into hypoxanthine in three steps via adenosine and inosine, with subsequent phosphoribosylation of hypoxanthine to IMP. The minor pathway involves formation of ATP, which is converted via the histidine pathway to the purine intermediate 5-amino-4-imidazolecarboxamide ribonucleotide and, subsequently, to IMP. Here we describe E. coli mutants, in which a third pathway for conversion of adenine to IMP has been activated. This pathway was shown to involve direct deamination of adenine to hypoxanthine by a manganese-dependent adenine deaminase encoded by a cryptic gene, yicP, which we propose be renamed ade. Insertion elements, located from −145 to +13 bp relative to the transcription start site, activated theade gene as did unlinked mutations in the hnsgene, encoding the histone-like protein H-NS. Gene fusion analysis indicated that ade transcription is repressed more than 10-fold by H-NS and that a region of 231 bp including theade promoter is sufficient for this regulation. The activating insertion elements essentially eliminated the H-NS-mediated silencing, and stimulated ade gene expression 2–3-fold independently of the H-NS protein.

Escherichia coli auxotrophic mutants can utilize adenine as the sole source of purines. Conversion of adenine into guanine nucleotides occurs by two different pathways that converge on IMP and utilize the subsequent reactions of the de novo synthesis pathway to guanine nucleotides (1). The major pathway involves conversion of adenine to hypoxanthine in three steps involving the intermediate formation of adenosine and inosine (Fig. 1). The first and third reactions in this sequence are catalyzed by the deoD gene product, purine nucleoside phosphorylase, whereas the second step is catalyzed by adenosine deaminase encoded by the add gene. Hypoxanthine in turn is converted to IMP by hypoxanthine or guanine phosphoribosyltransferases encoded by the hpt and gpt genes, respectively.
The minor pathway involves formation of ATP, which is converted via the histidine pathway to the purine intermediate 5-amino-4-imidazolecarboxamide ribonucleotide and subsequently to IMP (Fig. 1). The flux through this pathway is limited because the first enzyme of the histidine pathway, HisG, is subject to strong feedback inhibition by the end product histidine (2). Thus, purine-requiring mutants in which the major pathway is blocked by mutation of the deoD gene only grow very slowly with adenine as the sole source of purines, and this residual growth can be eliminated by the addition of histidine to the growth medium (3). These findings indicate that there are no other pathways for converting adenine into guanine nucleotides. Specifically, studies of enzymatic activities in crude extracts indicated that E. coli contains no adenine deaminase activity, which might convert adenine directly to hypoxanthine by deamination (4). Nevertheless, Kocharyan et al. (5) reported the isolation of E. coli mutants, in which an apparently cryptic adenine deaminase gene had been activated. The genetic locus affected in these mutants, however, was not identified nor mapped.
The paradigm example of cryptic genes in E. coli is the bgl operon involved in metabolism of aromatic ␤-glucosides. A key element in the silencing of the bgl operon is the small abundant nucleoid-associated protein, H-NS, which probably forms a repressing nucleoprotein complex upon binding to silencer DNA regions flanking the bgl promoter (6 -10). In addition to the bgl operon and other cryptic genes the H-NS protein also modulates the expression of a large number of active E. coli genes, usually by repressing transcription initiation (11,12). H-NS binds to DNA with no obvious sequence specificity, but specific binding sites tend to be AT-rich and intrinsically bend (13)(14)(15)(16), as also observed for the upstream silencing region of the bgl operon (6,8,17). Binding of H-NS generally induces strong condensation of DNA, and thus, the protein has been implicated in the organization and compaction of the bacterial nucleoid (18). H-NS consists of an N-terminal oligomerization domain and a C-terminal DNA binding domain, and the ability of the protein to condense DNA and repress transcription apparently depends on its ability to oligomerize (19 -21).
In the present work we have isolated and characterized mutants with increased adenine deaminase activity and demonstrate that this activity is due to a manganese-dependent adenine deaminase encoded by a cryptic gene, yicP, which we propose be renamed ade. In agreement with these findings, purified YicP protein has recently been shown to posses significant adenine deaminase activity (22). To define the elements responsible for the cryptic nature of the ade gene, we have identified a large number of cis-and trans-acting mutations that lead to activation of gene expression. Like the bgl promoter, the ade promoter region was found to be extremely AT-rich and subject to strong repression by the H-NS protein.
As also observed for the bgl operon, we found that insertion of a variety of IS elements within an extended region surrounding the ade promoter resulted in relief of the H-NS-mediated silencing. The results suggest that these IS 1 elements interfered with the formation of an H-NS⅐DNA complex, which would otherwise sequester the adjacent ade promoter region.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Media-The bacterial strains used in this study are all derivatives of E. coli K12 and are listed in Table I. Generalized transductions with lysates of bacteriophage P1vir were performed as described (26). Minimal medium plates contained AB minimal medium (27) solidified with 2% of Difco Bacto agar and supplemented with 0.2% glucose or glycerol as the carbon source, 1 g/ml thiamine, 15 g/ml nucleobases, 30 g/ml nucleosides, or 40 g/ml methionine and histidine when required. The rich medium was Luria broth (LB). Liquid cultures were grown in glucose minimal medium supplemented with 15 g/ml hypoxanthine and 0.3% casamino acids unless otherwise indicated.
Selection of Mutants with Increased Adenine Deaminase Activity-An auxotrophic purE deoD strain, CN1980 (Table I), was plated on glucose minimal medium containing 40 g/ml histidine and 15 g/ml adenine as the sole source of purines. Mutants that were still auxotrophic and capable of utilizing adenine as a purine source were characterized further as described under "Results." Several independent selections were performed. To facilitate subsequent analysis of the mutants obtained, some of these selections were performed on strains that were isogenic with CN1980 except that they contained gsk mutant alleles which did not give rise to kanamycin resistance. Verification of genomic mutations by colony PCR amplification and DNA sequencing was performed as described previously (23).
Isolation of an ade::cam Disruption Mutant, CN2451-A phage -sensitive LamB ϩ derivative of CN2388 -2 (Table I) was mutagenized with mini-Tn10 cam from NK1324 as described (28). After penicillin enrichment (23), we isolated chloramphenicol-resistant clones that had lost the ability to utilize adenine as a purine source by replica plating onto glucose minimal plates containing either adenine or hypoxanthine as a purine source. One of these mutants, CN2451, gave rise to a PCR product of ϳ1.5 kilobases with the cam-down and ade-Bam primers (Table II), and sequencing of this DNA fragment showed that the cam gene was located at positions 4156 -4164 2 within the ade gene in the same orientation as ade.
Plasmid Constructions-DNA manipulations, transformations, and restriction analyses were performed according to standard procedures (29). PCR amplifications were performed on 1 g of genomic DNA using Pfu polymerase (Stratagene) according to the manufacturer's recommendations. The DNA oligonucleotides used as primers in PCR reactions are described in Table II. The sequence of cloned PCR fragments were verified by sequencing on an ABI377 DNA sequencer (Applied Biosystems/PerkinElmer Life Sciences).
For construction of pAde, the ade gene and flanking regions (nucleotides 3612-5598) were PCR-amplified from genomic DNA with the ade-R1 and ade-Bam primers (Table II) and inserted between the EcoRI and BamHI sites of the medium-copy vector pBR322 (30).
For construction of medium-copy ade-lacZ gene fusions, the ade promoter and 65 base pairs of N-terminal coding region was PCR-amplified from genomic DNA and inserted between unique EcoRI and HindIII sites in the lacZ gene fusion vector pCN302, which is based on the pBR322 replicon (24). The ade promoter fragments in pCN2421 and pCN2422 were amplified with the ade-R1 and ade-H3 primers (Table II), whereas the promoter insert in pCN2534 was generated with the ade-R12 and ade-H3 primers (Table II). A low-copy derivative, pCN2515, of the IS1activated ade-lacZ fusion, pCN2422, was constructed by inserting the ade promoter fragment into the low-copy lacZ gene fusion vector, pCN2423, which is based on the pSC101 replicon (23).
Assays of Adenine Deaminase Activity in Crude Extracts-30 ml of bacterial culture was harvested on ice at A 436 ϭ 0.5. Cells were collected by centrifugation, washed, and resuspended in 50 mM Tris-HCl, pH 7.5, to A 436 ϭ 20. Cells were disrupted by sonic treatment, and the extract was cleared by centrifugation at 20,000 ϫ g for 3 min in a refrigerated microcentrifuge. Assays were performed at 37°C by mixing appropriately diluted extract with [8-14 C]adenine (8.4 Ci/mol) at a final concentration of 0.5 mM in a total volume of 50 l of 40 mM Tris-HCl, pH 7.5, 5 mM MnCl 2 . At timed intervals, samples of 15 l were taken out, boiled for 2 min, and cooled on ice. After a 3-min centrifugation at 20,000 ϫ g, 5 l of supernatant was applied to a polyethyleneimine thin layer chromatography plate. Substrate and products were separated by chromatography in water and subsequently quantitated by counting in an Instant Imager (Packard Instrument Co.). The enzymatic activities were calculated from the initial slope of a plot of the amount of radioactive substrate remaining as a function of time (see Fig. 2). The reported activities are the averages of two independent determinations, which deviated less than 10% from the average. The adenine deaminase was strongly dependent on manganese ions; the enzymatic activity decreased more than 30-fold if MnCl 2 was omitted from the assay (data not shown).
Measurements of ␤-Galactosidase Synthesis-Differential rates of ␤-galactosidase synthesis were measured at 37°C as described previously (31).
RNA Isolation and Primer Extension Analysis-Bacterial RNA was prepared by hot phenol extraction, and primer extension analysis was performed on 5 g of total RNA as described (32). We used a 32 P-labeled DNA primer complementary to nucleotides 41-64 of the lacZ coding sequence (#1224, see Table II).

Selection of Mutants with Increased Conversion of Adenine to
Guanine Nucleotides-Mutants with an activated adenine deaminase gene were sought by plating a purine-requiring deoD mutant strain, such as CN1980 (Table I), on glucose minimal medium containing adenine and histidine (see "Experimental Procedures"). With a frequency of 10 Ϫ6 , mutant colonies appeared that apparently converted adenine efficiently into guanine nucleotides despite the blocked deoD pathway and the presence of histidine in the medium. Sixty of such independent mutants were characterized further and found to belong to three distinct classes as described in the following.
Class I; Mutants with a Deregulated Histidine Biosynthetic Pathway-Half of the isolated mutants lost the ability to grow with adenine as a purine source upon introduction of the hisG::Tn10 allele from NK5526 (Table I), and for several of these mutants the responsible mutation was mapped to the hisG gene itself (data not shown). Thus, the feedback regulation of the HisG enzyme was probably eliminated in this class of mutants to allow an increased flux through the histidine biosynthetic pathway even in the presence of histidine. These mutants were not characterized further.
Class II; Mutants with Increased Adenine Deaminase Activity Caused by Mutations Closely Linked to the yicP Gene-The remaining mutants did not lose the ability to grow on adenine upon introduction of a hisG::Tn10 or a deoD::cam mutation (data not shown), suggesting that a novel third pathway for conversion of adenine into guanine nucleotides had been activated. Based on differences with respect to growth and colony morphology, these mutants could be divided into two distinct classes (II and III) consisting of 20 and 10 independent clones, respectively. The Class II mutants, like the parent strain, grew well on a variety of carbon sources and formed normal nonmucoid colonies. Genetic analysis revealed that the responsible mutation in the Class II mutants showed a high cotransduction frequency with the ilvB locus (data not shown), which is compatible with a location in the immediate vicinity of the yicP gene. Thus, we hypothesized that the novel activated pathway in Class II mutants might involve direct deamination of adenine to hypoxanthine by the yicP gene product.
In support of this notion, we found that a crude extract of one of the Class II mutants, CN2388 -2, efficiently converted adenine into hypoxanthine with no apparent formation of any intermediate products, corresponding to a 20-fold increase of the cellular adenine deaminase activity compared with the parent strain, CN1980 (Fig. 2, left and center panels). Furthermore, we subjected a derivative of CN2388 -2 to transposon mutagenesis with mini-Tn10 cam and isolated a clone that had specifically lost the ability to grow with adenine as a purine source while retaining the ability to utilize hypoxanthine. Subsequent analysis revealed that the cam insert in this strain, CN2451, had indeed disrupted the yicP gene (see "Experimental Procedures"). Back transduction of the yicP::cam allele into all the Class II mutants eliminated their ability to use adenine as a purine source (data not shown), confirming that the original phenotype was caused by activation of the yicP gene, which we propose be renamed ade.
This conclusion was further corroborated by cloning of the chromosomal ade gene and its native promoter in a multicopy plasmid, pAde (see "Experimental Procedures"). Introduction of this plasmid increased the adenine deaminase activity of CN1980 ϳ30-fold from 0.5 to 15.4 mol/min/g of dry weight and enabled the transformant to grow well in glucose minimal medium with adenine as the sole source of purines at a rate comparable with that of a purine-prototrophic strain (data not shown). These results indicated that the wild type ade gene encodes a functional adenine deaminase but is simply too poorly expressed at normal gene dosage to satisfy the cellular purine requirement, which is on the order of 4 mol/min/g of dry weight (33). Indeed, the low but significant adenine deaminase activity in the wild type strain, CN1980, was completely abolished by the ade::cam disruption in CN2481 (compare Figs. 2, left and right panels). In agreement with the results of Matsui et al. (22), the E. coli adenine deaminase was strongly dependent on manganese ions for activity (see "Experimental Procedures"), as also observed for the homologous enzyme from B. subtilis (36).

Nature of the Activating Mutations in the Class II Mutants-
The mutations responsible for the increased adenine deaminase activity were identified for 15 Class II mutants by PCR amplification and DNA sequencing of the ade promoter region (Fig. 3). In 12 of these mutants the ade gene was activated by integration of an IS1 element within the first 100 base pairs of the divergent neighboring gene, yicO. Two mutants were activated by insertion of IS4 or IS5 in the same region of yicO, whereas one mutant contained an IS5 insertion in the intercistronic region between ade and yicO. The latter insertion was the only one found to be located downstream of the ade promoter. In addition we found two Class II mutants with unidentified insertions or rearrangements in yicO that could not be spanned by PCR amplification as well as three mutants that apparently contained an amplification of the entire ade region including the ilvB locus (data not shown).
The finding that all the identified Class II mutations mapped outside of the structural ade gene underscored that the cryptic nature of this locus was caused by a low level of expression rather than by malfunctioning of the encoded gene product. It is also noteworthy that activation could occur by insertion of IS elements within an extended region of 160 base pairs surrounding the ade promoter and that this activation, at least for IS1, was independent of the orientation of the insertion element.
Class III; Mutants with Increased Expression of the ade Gene Caused by Mutations in the Unlinked hns Gene-The Class III mutants were clearly distinguishable from the parent strain and the other mutant classes by their formation of mucoid colonies even on LB plates and their poor growth with glycerol or succinate as carbon sources irrespective of the purine source. The cellular adenine deaminase activity of a representative Class III mutant, CN2388-16 (Table I), was found to be 3.8 mol/min/g of dry weight, which is nearly 8-fold higher than the enzymatic activity of the wild type parent. Furthermore, CN2388-16 lost the ability to grow with adenine as a purine source upon introduction of the ade::cam allele. These results indicated that the Class III mutants, like those of Class II, contain an activated ade gene. However, transductional mapping experiments indicated that the responsible activating mutation was unlinked to the ade and ilvB loci (data not shown), suggesting a mutation in a trans-acting regulatory locus.
Accordingly, we mapped the mutation responsible for the increased adenine deaminase activity and the other phenotypes of CN2388-16 to the 27-min region of the genome immediately clockwise of the supF marker (data not shown). This corresponds to the location of the hns gene, mutations in which are known to cause increased formation of capsular polysaccharides and poor growth on gluconeogenic carbon sources (37). Thus, we inferred that the Class III mutants might contain loss-of-function mutations in the hns gene, and this was confirmed by PCR amplification of the hns gene in CN2388-16 and another Class III mutant, CN2388 -11, using the primers hns-RV ϩ hns-H3 (Table II). DNA sequencing of the PCR fragments revealed that the hns gene in both cases had been disrupted by insertion of an IS1 element in the region encoding the C-terminal DNA binding domain of H-NS, 2 leaving a truncated reading frame of 126 or 108 codons, respectively (plus additional codons encoded by the IS1 sequence).
In agreement with these results, we found that the cellular adenine deaminase activity increased 11-fold upon introduction of the well characterized hns205::Tn10 allele into CN1980 (Table III). This hns disruption allele, which codes for a truncated H-NS protein of only 93 amino acid residues (38), also mimicked the other phenotypes of the selected Class III mutants with respect to increased mucoidicity and poor growth on glycerol or succinate. All these results indicated that the cryptic nature of the ade gene in wild type strains is due to H-NSmediated gene silencing.

IS Insertions Antagonize H-NS-mediated Repression of the ade Gene-To investigate if the IS insertions of the Class II
mutants activated the ade gene independently of the H-NSmediated regulation, the hns205::Tn10 allele was transduced into three of the Class II mutants, CN2388 -2 and CN2388 -15, which are activated by an IS1 insertion, and CN2388 -6, which is activated by IS4. In the hns ϩ background, the insertions in these strains gave rise to a dramatic increase of the cellular adenine deaminase activity, ranging from 10-to 27-fold (Table  III). However, the enzymatic activity produced from these alleles only increased by an additional 1.4 -2.4-fold upon introduction of the hns205::Tn10 allele, which should be contrasted with the more than 10-fold stimulation seen for the wild type ade gene upon disruption of hns in CN1980 (Table III). Thus, Apparently, the IS insertions also increased the intrinsic expression of the ade gene in the absence of H-NS, as seen by the 2-3-fold higher enzymatic activities compared with the wild type ade allele in the hns mutant background (Table III). Moreover, the ability of the IS insertions to prevent H-NSmediated repression seemed to correlate with their intrinsic stimulating effect on the ade gene, which might suggest that the two effects are mechanistically related.
A 231-bp Region Including the ade Promoter Is Sufficient to Mimic the Silencing and Activation of the Intact ade Gene-To define the target for the H-NS-mediated regulation of the ade gene, we measured the adenine deaminase activity produced from pAde in a wild type strain, CN2349, and its isogenic hns::Tn10 derivative, CN2498. As shown in Table III expression from pAde was repressed more than 10-fold by H-NS, as also observed for the chromosomal ade gene in CN2349 (transformed with the vector plasmid pBR322). This corresponds closely to the 11-fold regulation observed previously for the chromosomal ade gene in the very different CN1980 background (Table III). Thus, the region cloned on pAde, extending from the ade-R1 primer region (Fig. 3) to immediately downstream of the ade reading frame is sufficient to confer H-NSmediated repression.
The target for the H-NS-mediated regulation was further defined by construction and characterization of ade-lacZ gene fusions (Fig. 4). The data for pCN2421 showed that the 231base pair-long region bounded by the ade-R1 and the ade-H3 primer targets (Fig. 3) was sufficient to give more than 10-fold H-NS-mediated regulation of the ade-lacZ fusion. The inclusion of an additional 853 base pairs of yicO sequence upstream of the ade promoter in pCN2534 did not significantly affect the regulation by H-NS but reduced gene expression 2-fold both in the wild type and hns mutant background. These results corroborated the finding for pAde (Table III) that sequences upstream of the ade-R1 primer region were not specifically required for H-NS-mediated repression and further established that sequences downstream of codon 22 in the ade gene were also dispensable for this regulation.
In agreement with these results, we found that the yicO-2::IS1 insertion stimulated expression of the ade-lacZ fusion in pCN2422 to the same extent as it stimulated expression of the chromosomal ade gene in CN2388 -2 (compare Fig. 4 and Table III). The presence of the IS element almost eliminated the H-NS-mediated regulation of the lacZ fusion to a mere 1.4-fold and apparently increased the intrinsic strength of the ade promoter as seen from the 3-fold higher expression of pCN2422 compared with pCN2421 in the hns mutant background. The virtual abolition of H-NS-mediated repression of pCN2422 was not an artifact caused by the very high expression level of this construct. A low copy derivative of this plasmid, pCN2515, was similarly unresponsive to the hns mutation (Fig. 4). It might be imagined that the activating effect of the IS insertions derived from disruption of the divergent yicO gene if the YicO protein was somehow required for silencing of the ade gene. However, pCN2422 contains only the first 23 codons of yicO, and yet, the yicO::IS1 disruption in this plasmid had a similar activating effect as in the chromosomal context. This result strongly suggests that the IS insertions activated the neighboring ade promoter by a cis-effect independently of the YicO protein.
Finally, it should be noted that the stimulating effect of the hns mutation on expression of the ade-lacZ fusions was not caused by an unspecific effect on plasmid copy numbers or on   (Table II)  ␤-galactosidase synthesis as such. Expression of the lacZ gene both in the medium-copy vector, pCN302, and in the low-copy vector, pC2423, was actually reduced by ϳ30% in the hns background (data not shown).
Effect of an IS Insertion and the H-NS Protein on Transcription from the ade Promoter-To investigate how the IS sequences and the H-NS protein modulate ade gene expression, we mapped by primer-extension analysis the 5Ј-ends of ade-lacZ fusion transcripts produced in a wild type or a hns mutant background (Fig. 5). In the wild type background the unactivated fusions pCN2421 and pCN2534 only gave rise to a very faint signal that was hardly visible on the reproduction in Fig.  5 (but clearly visible on longer exposures). However, the same signal increased greatly in strength in the hns mutant background, in close agreement with the ␤-galactosidase synthesis from these constructs (Fig. 4). This signal corresponded to a major 5Ј-end located 54 nucleotides upstream of the ade initiation codon, it was preceded by likely Ϫ10 and Ϫ35 promoter signals, and thus, probably represented the ade promoter (Fig.  3). The ade-lacZ fusion with the activating IS insertion upstream of the ade promoter pCN2515 showed a strong signal at exactly the same position even in the wild type background, and this signal was only marginally stimulated by the hns mutation (Fig. 5), again in close agreement with the ␤-galactosidase activities measured for this construct (Fig. 4).
These results indicate that the absence of the H-NS protein or the presence of the activating IS1 insertion caused a marked derepression of the ordinary ade promoter rather than promoting transcription from alternative start sites. In particular there was no indication that transcription originated from within the activating IS element in pCN2515 (pCN2422). In this connection it is intriguing that some of the IS1 insertions were located 83-120 nucleotides upstream of the Ϫ35 signal (Fig. 3), which is even upstream of the region that was sufficient for H-NS-mediated silencing of the ade gene (Fig. 4). These findings suggest that the IS1 insertions somehow modulated the DNA structure in the downstream ade promoter region to stimulate transcription and interfere with H-NS-mediated silencing.

DISCUSSION
Physiological Role of the E. coli ade Gene-Although the ade gene is cryptic in E. coli, the present work demonstrates that the encoded protein is a fully functional manganese-dependent FIG. 5. Primer extension analysis of ade-lacZ fusion transcripts produced in a wild type or a hns mutant background. RNA was harvested from a wild type strain, CN2349 (ϩ), or its isogenic hns205::Tn10 derivative, CN2498 (Ϫ) transformed with the indicated plasmids. The primer extension reactions were loaded on an 8% sequencing gel together with a sequencing ladder made with the same 32 P-labeled primer on the ade-lacZ fusion plasmid, pCN2421. The position of the major mRNA 5Ј-end is indicated on the sequence in Fig. 3.  4. Effect of the hns205::Tn10 mutation on expression of ade-lacZ gene fusions. Differential rates of ␤galactosidase synthesis were measured for the indicated plasmids transformed into a wild type strain and its isogenic hns205::Tn10 derivative (CN2349 and CN2498, respectively, Table I). The extent of the ade promoter region that is fused to lacZ in each construct is indicated by a black bar. The promoter fragments in pCN2421, pCN2422, and pCN2515 are all bounded by the ade-R1 and ade-H3 primer targets (Fig. 3), whereas pC2534 contains an additional 853 base pairs of upstream yicO sequence. adenine deaminase, in agreement with the recent findings of Matsui et al. (22). This enzyme is homologous (34% identity) to the adenine deaminase of Bacillus subtilis (36), and homologues of these proteins are encoded in the genomes of many bacteria and Archaea, but apparently not in eukaryotic genomes according to the Pfam data base (Pfam01979, Ref. 39) An adenine deaminase gene, AAH1, has been identified in Saccharomyces cerevisiae (40), but according to sequence homology, the encoded protein belongs to the family of deaminases acting on adenosine or AMP (Pfam00962, Ref. 39).
The physiological role of the ade gene in wild type E. coli cells is obscure. Considering the cryptic nature of the gene and the apparent redundancy with the deoD add pathway for conversion of adenine to hypoxanthine, it is hardly surprising that disruption of the ade gene had no obvious phenotypic consequences. The finding that the structural gene has remained functional during evolution, on the other hand, suggests that it may be induced and have a beneficial role under some physiological conditions. For instance the adenine deaminase might contribute to the utilization of adenine as a purine or nitrogen source under conditions where function of the deoD pathway is restricted. In this connection it is appealing that the divergent yicO gene, according to sequence homology, may code for a purine nucleobase permease, especially as the probable overlap of the two divergent promoters suggests that they might be co-regulated. It should be emphasized, however, that the yicO gene product was dispensable for adenine uptake and seemed to play no role in silencing of the ade promoter.
So far we have not uncovered physiological conditions that lead to activation of the ade locus. The inability of pur deoD mutants to grow with adenine as a purine source (at 15 g/ml) shows that the gene is not sufficiently induced by purine starvation or by the adenine substrate itself. However, Matsui et al. (22) report that growth in the presence of a very high concentration of adenine (200 g/ml) resulted in a slight (15-50%) increase of the cellular adenine deaminase activity. Whether this reflects a specific regulation of the ade gene by adenine is presently unclear. We found that both cell growth and expression of the ade-lacZ fusions were somewhat inhibited at such high adenine concentrations (data not shown).
Unfortunately, the finding of H-NS-mediated regulation helps little with respect to identifying physiological inducing conditions for the ade gene, because conditions leading to general alleviation of H-NS-mediated repression are unknown. The cryptic bgl operon is activated by point mutations that improve the CRP binding site of the bgl promoter (6), and several other H-NS repressed genes are activated by specific DNA-binding proteins that antagonize H-NSmediated repression in response to specific stimuli (41)(42)(43)(44)(45). The region upstream of the Ϫ35 signal of the ade promoter resembles a CRP binding site (Fig. 3), but expression of the ade-lacZ fusions were not markedly reduced in a strain devoid of the CRP (data not shown), and our genetic selections have not implicated the existence of any other regulatory protein specific for the ade locus.
H-NS-mediated Silencing of the ade Locus-In the present work we found that mutations truncating the C-terminal DNA binding domain of H-NS strongly induced ade gene expression. In marked contrast, the bgl operon is not markedly activated by mutations affecting the C-terminal domain of H-NS, including the hns205::Tn10 allele used here (19,38). Apparently, repression of the bgl operon is retained in such mutants because the truncated H-NS protein forms a complex with a homologous protein, StpA, which may serve as an adapter, providing a DNA binding domain to the N-terminal oligomerization domain of H-NS (38). Our results indicate that this kind of complemen-tation does not work efficiently in silencing of the ade gene, as previously observed for H-NS mediated repression of the proU gene (19).
A region of only 231 bp surrounding the ade promoter was sufficient for H-NS-mediated repression of the ade promoter, but we have not directly demonstrated a physical interaction between the H-NS protein and this region. In DNase I footprinting experiments, H-NS typically protects extended target regions of 100 base pairs or more, which are generally AT-rich and curved (11). We think it is most likely that the ade gene is silenced by a direct interaction between H-NS and the ade regulatory region, because the entire region implicated in silencing is extremely AT-rich (75%) and contains numerous poly-dA and poly-dT tracts that may impose intrinsic curvature on the DNA (Fig. 3). Notably these features are also shared by the regulatory regions of several other genes that are silenced by H-NS (6,(45)(46)(47).
So far we have obtained no DNA gyrase mutants in our selections for strains with increased adenine deaminase activity, and expression of the ade-lacZ fusion was not stimulated in a strain deficient in DNA gyrase activity (data not shown). These findings suggest that silencing of the ade promoter is independent of the general level of DNA supercoiling, which is compatible with previous findings that H-NS binding generally is unaffected by the superhelical status of the DNA (18,48,49). On the other hand, the bgl operon is known to be activated under conditions of reduced DNA supercoiling, but this effect seems to be specifically related to the presence of an extended dyad symmetry in the upstream silencer region, which may extrude into a cruciform structure in negatively supercoiled DNA (9).
Activation by IS Elements-The ade gene was strongly activated by insertion of a variety of IS elements, (including IS1, IS4, IS5, and others) into a region extending from Ϫ145 to ϩ13 relative to the transcription start site of the ade promoter ( Fig.  3 and Table III). This phenomenon of "transpositional activation" has also been observed for the bgl operon, which may be activated by insertions of predominantly IS1 or IS5 into a region extending from Ϫ132 to ϩ91 relative to the transcription start site (6,50). In analogy with the previous studies of the bgl system, our results indicated that the IS1 elements activated the dormant ade promoter in an orientation-independent manner mainly by interfering with H-NS-mediated repression. Interestingly, some of the activating IS elements were even located upstream of the region that appeared to be sufficient for H-NS-mediated repression.
It is not completely clear how insertion elements may prevent repression of a downstream promoter, but Schnetz and Rak (50) find that activation of the bgl promoter by IS5 is at least partially dependent on the IS5-encoded transposase protein, which binds to the ends of the IS5 element and is functional in activation even if delivered in trans. Furthermore, DNA-binding proteins, which bind in the vicinity of H-NS repressed promoters, are frequently observed to alleviate repression (6,(41)(42)(43)(44)(45). Taken together, all these findings suggests that IS elements activate transcription by providing binding sites for DNA-binding proteins, which distort the DNA topology in the adjacent H-NS target region and thereby prevent the formation of a silencing H-NS-DNA nucleoprotein structure.
We hypothesized that the abundant nucleoid binding protein, integration host factor, might be involved in the activating effect of the IS1 element, since it is known to bind strongly to the ends of the IS1 element and bend the DNA (51). However, silencing and activation of the ade-lacZ fusions were unaffected by a disruption of the himA gene, which encodes one of the two integration host factor subunits (data not shown). Thus, if protein binding is involved in IS1-mediated activation, it seems more likely that the crucial protein is the IS1-encoded transposase, in analogy with IS5-mediated activation.
Interestingly, the IS1 and IS4 elements tested here also seemed to stimulate transcription from the ade promoter independently of the H-NS protein (Table III). Conversely, the upstream yicO region between the ade-R12 and ade-R1 target sites negatively affected ade-lacZ expression independently of H-NS (Fig. 4). These findings suggest that the very AT-rich ade promoter region was particularly sensitive to changes in DNA topology imposed by neighboring DNA sequences, perhaps because it was particularly prone to unwinding or bending.