Purine Utilization by Klebsiella oxytoca M5al: Genes for Ring-Oxidizing and -Opening Enzymes

Abstract

The enterobacterium Klebsiella oxytoca uses a variety of inorganic and organic nitrogen sources, including purines, nitrogen-rich compounds that are widespread in the biosphere. We have identified a 23-gene cluster that encodes the enzymes for utilizing purines as the sole nitrogen source. Growth and complementation tests with insertion mutants, combined with sequence comparisons, reveal functions for the products of these genes. Here, we report our characterization of 12 genes, one encoding guanine deaminase and the others encoding enzymes for converting (hypo)xanthine to allantoate. Conventionally, xanthine dehydrogenase, a broadly distributed molybdoflavoenzyme, catalyzes sequential hydroxylation reactions to convert hypoxanthine via xanthine to urate. Our results show that these reactions in K. oxytoca are catalyzed by a two-component oxygenase (HpxE-HpxD enzyme) homologous to Rieske nonheme iron aromatic-ring-hydroxylating systems, such as phthalate dioxygenase. Our results also reveal previously undescribed enzymes involved in urate oxidation to allantoin, catalyzed by a flavoprotein monooxygenase (HpxO enzyme), and in allantoin conversion to allantoate, which involves allantoin racemase (HpxA enzyme). The pathway also includes the recently described PuuE allantoinase (HpxB enzyme). The HpxE-HpxD and HpxO enzymes were discovered independently by de la Riva et al. (L. de la Riva, J. Badia, J. Aguilar, R. A. Bender, and L. Baldoma, J. Bacteriol. 190:7892-7903, 2008). Thus, several enzymes in this K. oxytoca purine utilization pathway differ from those in other microorganisms. Isofunctional homologs of these enzymes apparently are encoded by other species, including Acinetobacter, Burkholderia, Pseudomonas, Saccharomyces, and Xanthomonas.

Purines and purine derivatives comprise a large portion of biomass and are involved in almost every step of life. Not only a major constituent of nucleic acids, they also are central to energy transfer and storage (ATP) as well as protein synthesis and signaling (GTP). Plants, animals, and many microorganisms use purines and purine derivatives to store and translocate nitrogen for assimilation or excretion (96).

Salvage pathways operate to recycle purines, including hypoxanthine and xanthine, back into nucleoside pools (107). Additionally, some organisms can utilize purines as the sole source of nitrogen and carbon. Adenine and guanine are deaminated to form hypoxanthine and xanthine, respectively, which then are oxidized to form uric acid (urate at physiological pH) (Fig. 1). These oxidation steps are catalyzed by xanthine dehydrogenase, a well-studied molybdoflavoenzyme that is conserved from bacteria to humans (51). Two sequential ring-opening steps convert urate via allantoin to allantoate (Fig. 1). Subsequent steps, which comprise different pathways in different microorganisms (96), convert allantoate to ammonium, which is assimilated.

FIG. 1.

Purine ring oxidation and opening steps. The enzyme proposed to catalyze each step is shown. The K. oxytoca gene for adenine deaminase was not identified in this study. Dashed lines show reactions that can occur spontaneously.

Some organisms express only the latter portion of the purine utilization pathway and cannot use purines or urate as sole sources of nitrogen. For example, Escherichia coli K-12 can use allantoin and its catabolites as the sole nitrogen source, albeit only under anaerobic conditions (21). Saccharomyces cerevisiae uses allantoin as a nitrogen storage compound (17). However, the complete pathway is present in other bacterial and fungal species, including Bacillus subtilis (84) and Aspergillus nidulans (83).

Molybdoenzymes (excepting dinitrogenase) contain the molybdenum cofactor Mo-molybdopterin (42). Thus, mutations in genes for molybdenum cofactor biosynthetic enzymes (mol genes in bacteria and cnx in A. nidulans) confer pleiotropic phenotypes: these mutants can utilize neither nitrate nor purines, due to lack of the molybdoenzymes nitrate reductase and xanthine dehydrogenase (74). We previously reported that Klebsiella oxytoca mol mutants cannot assimilate nitrate but can utilize xanthine as the sole nitrogen source (32). This suggested, as one possibility, that K. oxytoca uses a molybdenum-independent enzyme in place of conventional xanthine dehydrogenase. Results reported here demonstrate that this is correct, as insertion mutants blocked specifically in xanthine and hypoxanthine utilization define the structural genes for an apparent two-component Reiske nonheme iron oxygenase.

Here, we report analysis of 12 genes whose products catalyze conversion of purines to allantoate. Our investigation of the remaining genes, whose products catalyze allantoate utilization, is ongoing. Results show that several steps in the overall pathway are catalyzed by previously undescribed enzymes.

While this paper was in review, the paper by de la Riva et al. (24), describing the hpxDE, hpxR, hpxO, and hpxPQT genes from Klebsiella pneumoniae W70, was posted in the “JB Accepts” section of the Journal of Bacteriology online edition. Results and conclusions concerning these seven genes are congruent between the two studies.

(Some of the work presented here was submitted by Danielle Carl in 1994 as part of an undergraduate thesis to the Cornell University Division of Biological Sciences Honors Program.)

MATERIALS AND METHODS

Media.

Defined, complex, and indicator media for routine genetic manipulations were used as described previously (57). Nitrogen-free medium contained 0.2% (wt/vol) glucose, 1% (wt/vol) sodium citrate, 0.74% (wt/vol) sodium phosphate (pH 8), and 1 mM MgSO₄ (53). This medium was supplemented with additional nitrogen sources (10 mM NaNO₃ or NH₄Cl and 2.5 mM hypoxanthine, xanthine, urate, allantoin, or potassium allantoate) as indicated. Urate solutions were prepared as described previously (79). 5-Bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal; 40 μg/ml) was included at 40 μg ml⁻¹ to score the Lac phenotypes of MudJ insertion mutants. All nitrogen sources were from Sigma-Aldrich (St. Louis, MO).

Selection for K. oxytoca transformants carrying bla-containing plasmids was accomplished with a combination of carbenicillin and ampicillin (Ap) at 800 and 60 μg ml⁻¹, respectively (54). Streptomycin (Sm) was used at 500 μg ml⁻¹ for selecting K. oxytoca Sm^r segregants (102). Other antibiotics used were chloramphenicol (Cm) at 50 μg ml⁻¹, kanamycin (Km) at 100 μg ml⁻¹, spectinomycin (Sp) at 50 μg ml⁻¹, and tetracycline (Tc) at 20 μg ml⁻¹.

Defined medium used to grow cultures for β-galactosidase assays and for growth yield tests was buffered with 3-[N-morpholino]propanesulfonic acid (MOPS) as previously described (69). Glucose (40 mM) was used as the sole carbon source, and nitrogen sources were added as indicated.

Culture conditions.

Liquid cultures and plates were incubated at 30°C (41). Cultures for β-galactosidase assays were aerated at 240 rpm in 10 ml of medium in 125-ml sidearm flasks. Culture densities were monitored with a Klett-Summerson photoelectric colorimeter (Klett Manufacturing Co., New York, NY) equipped with a no. 66 (red) filter. Cultures in the mid-exponential phase (about 40 Klett units) were harvested, chilled on ice, and washed with saline. Cell pellets were stored overnight at −20°C prior to the assay.

Cultures for nitrogen assimilation tests were grown in 5 ml of MOPS medium in 16- by 175-mm culture tubes, aerated on a roller drum. Nitrogen sources were added over a range of growth-limiting concentrations: for nitrate, 0.4 to 1.8 mM; for histidine, 0.2 to 0.9 mM; for xanthine, 0.1 to 0.45 mM; and for guanine, 0.08 to 0.36 mM. Culture densities (optical densities at 420 nm) were measured after 24 and 48 h.

Phenotypes were determined by growth on plates made with nitrogen-free medium and microbiological agar (Marine BioProducts, Delta, British Columbia) supplemented with the test nitrogen source at a final concentration of 10 mM N atoms. Strains were first streaked for single colonies on plates with added nitrate, which then were replica printed to plates with test compounds (49). Plates for anaerobic nitrogen assimilation tests were incubated in Brewer-Allgeier jars (11).

β-Galactosidase assays.

Activity in CHCl₃-sodium dodecyl sulfate-permeabilized cells was measured at room temperature, approximately 21°C, by monitoring the hydrolysis of o-nitrophenyl-β-d-galactopyranoside. Specific activities are expressed in arbitrary (Miller) units (63). Cultures were assayed in duplicate, and reported values are averaged from two independent experiments.

Strains.

Table 1 lists the strains used in this study. Previously, strain M5al was classified under Aerobacter aerogenes and subsequently under K. pneumoniae. However, phenotypic properties (such as positive reaction in the indole test) place this strain in the species K. oxytoca (see reference 102). Genetic crosses were performed by P1 kc-mediated generalized transduction (90).

TABLE 1.

Strains and plasmids

Strain or plasmid	Genotype or description	Source or reference
K. oxytoca M5al strains
VJSK009	hsdR1	13
VJSK014	hsdR1 lac Tn7	13
VJSK038	As VJSK014 but mal::Mu cts hP1/pEG5005	13
VJSK650	As VJSK014 but hpxO102::Tn10d(Tc) mal::Mu cts hP1	This study
VJSK1371	As VJSK009 but mdr::Tn5	Laboratory collection
VJSK2088	As VJSK009 but lacZ101::Tn10d(Tc) rpsL	102
VJSK2966	As VJS2088 but ΔrhaBS::[Φ(hpxD-lacZ)]	This study
VJSK2967	As VJS2088 but ΔrhaBS::[Φ(hpxR-lacZ)]	This study
VJSK2968	As VJS2088 but ΔrhaBS::[Φ(hpxD-lacZ)] hpxR113::Ω-Cm	This study
VJSK2969	As VJS2088 but ΔrhaBS::[Φ(hpxR-lacZ)] hpxR113::Ω-Cm	This study
VJSK3046	As VJSK009 but hpxE108::MudJ zxx::Mu cts hP1	This study
VJSK3047	As VJSK009 but zxx::Mu cts hP1/pVJS3928	This study
Plasmidsa
pAH144	Smr; ori R6K; pUC18 multiple cloning site	39
pEG5005	Apr Kmr; ori pMB1; Mud5005	38
pHG165	Apr; ori pMB1; pUC8 multiple cloning site	89
pSU18	Cmr; ori p15A; pUC18 multiple cloning site	4
pKAS46	Apr Kmr; ori R6K; rpsL+; pSL1180 multiple cloning site	88
pVJS1435	As pHG165; (HindIII) 1284 to HindIII 17197 (hpxD-hpxU′)	This study
pVJS1436	As pHG165; (HindIII) 2344 to (HindIII) 12193 (hpxR-hpxB′)	This study
pVJS1454	As pSU18; SalI 11721 to (SalI) 23740 (′hpxA-hpxK)	This study
pVJS2354	As pKAS46 but ΔrhaBS::′lacZlacY+	103
pVJS3928	Apr Kms Spr; as pEG5005 but Ω-Sp	This study
pVJS3941	As pHG165; (HindIII) 1 to (HindIII) 8681 (hpxE-guaD)	This study
pVJS3958	As pHG165; HpaI 164 to HpaI 4180 (hpxE-′hpxO)	This study

^aSequence coordinates for restriction sites are shown. Sites in parentheses are from flanking Mu DNA present in the pEG5005 vector.

MudJ, originally termed Mu dI 1734, is a transposase-null bacteriophage transposon that encodes resistance to Km and is used for isolating lacZ operon fusions (14). Mu cts hP1, a P1 host range variant of bacteriophage Mu (19), was used as a helper phage for K. oxytoca as described previously (13). Tn10d(Tc) and Tn5d(Sp) are transposase-null transposons encoding resistance to Tc and Sp, respectively (25, 47).

Nomenclature.

The genetic nomenclature extends that used by de la Riva et al. (24). The mnemonic hpx (hypoxanthine) is applied to all genes whose products are involved in utilization of hypoxanthine or its catabolites.

We use Hpx as a general phenotypic designation without reference to a particular step in the pathway. Specific phenotypes are denoted according to the blocked step (Fig. 1); for example, Hxn⁻ mutants utilize neither hypoxanthine nor xanthine but grow normally with urate and downstream purine catabolites. Similarly, downstream blocks are denoted Urt⁻ (urate), Aln⁻ (allantoin), and Alt⁻ (allantoate).

Transposon insertions.

Table 2 lists insertion alleles used in this study. Insertion mutagenesis with transposons MudJ (13), Tn5d(Sp) (35), and Tn10d(Tc) (57) was performed as described previously. Insertions were selected on antibiotic plates, with nitrate as the sole nitrogen source, to preclude isolation of mutants with general defects in nitrogen assimilation. Colonies were then replica printed to antibiotic plates, with purine utilization intermediates (hypoxanthine, xanthine, urate, allantoin, or allantoate) as the sole nitrogen source.

TABLE 2.

hpx insertions

Allelea	Insertion coordinatesb	Product of codonc	Straind
hpxE108::MudJ (Lac+)	707-708	Gly-187	VJSK651
hpxE109::MudJ (Lac+)	739-740	His-176	VJSK652
hpxD103::MudJ (Lac+)	2337-2338	Nonee	VJSK643
hpxR113::Ω-Cm	2617-2618	Thr-60	VJSK2962f
hpxO102::Tn10d(Tc)	4079-4080	Ser-211	VJSK642
hpxO112::Tn5d(Sp)	4466-4467	Leu-82	VJSK3034
guaD114::pAH144	7602-7603	Leu-335	VJSK2985g
hpxA101::Tn10d(Tc)	11782-11783	Gln-242	VJSK641
hpxB104::MudJ (Lac−)	12052-12053	Tyr-79	VJSK644
hpxB107::MudJ (Lac−)	12189-12190	Asp-124	VJSK648
hpxB105::MudJ (Lac+)	12295-12296	Lys-160	VJSK645
hpxB110::MudJ (Lac−)	12515-12516	Thr-233	VJSK681
hpxF111::Tn5d(Sp)	17840-17841	Asp-28	VJSK3033

^aThe Lac phenotype for MudJ alleles is indicated in parentheses.

^bThe insertion or disruption lies between the indicated sequence coordinates.

^cCodon disrupted.

^dAll are in the VJSK009 strain background except as noted.

^eThe insertion lies 20 nt upstream of the hpxD initiation codon.

^fVJSK2088 strain background.

^gVJSK1371 strain background.

Plasmids.

Table 1 lists the plasmids used in this study. Standard methods (57) were used for restriction endonuclease digestion, ligation, transformation, and PCR amplification (80). Plasmid pVJS3982 was constructed by excising the Km^r determinant from plasmid pEG5005 with restriction endonuclease EcoRI and inserting the Ω-Sp integron (30) in its place.

Molecular cloning.

In vivo cloning utilized the mini-Mu cloning vectors pEG5005 and pVJS3928 as described previously (53). Lysates prepared from the hpx⁺ donor strains VJSK038 and VJSK3047 were used to transduce hpx insertion mutants lysogenic for Mu cts hP1. Strain VJSK650 (hpxO; Tc^r) was the recipient for cloning with vector pEG5005 (Km^r), and strain VJSK3046 (hpxE; Km^r) was the recipient for cloning with vector pVJS3982 (Sp^r). Antibiotic-resistant transformants were selected on complex medium, and colonies were replica printed to duplicate nitrogen-free plates, one containing nitrate and the other xanthine or hypoxanthine. Hpx⁺ colonies were isolated, and plasmid DNA was extracted and retransformed to confirm the phenotype.

This in vivo cloning method generates essentially random inserts whose boundaries are delimited by the ends of bacteriophage Mu DNA (38). Large inserts from the pEG5005-based clones were subcloned into the pHG165 or pSU18 medium-copy-number vector to generate the plasmids pVJS1435, pVJS1436, pVJS1454, and pVJS3941 (Table 1). Subcloning utilized HindIII or SalI restriction endonuclease sites near the Mu DNA ends. These sites are indicated in parentheses in Table 1. All other sites are native to the K. oxytoca hpx DNA region. Sites are denoted according to the nucleotide position within the sequenced region (see below).

DNA sequencing.

Automated DNA sequencing reactions (82) were performed by the DNA Sequencing Facility, University of California—Davis. Double-stranded templates were sequenced on a model 373A stretch DNA sequencer by using dye terminator chemistry and AmpliTaq-FS DNA polymerase (Perkin Elmer/Applied Biosystems Division, Foster City, CA). Templates were prepared by using QIAprep spin plasmid kits (Qiagen, Inc., Chatsworth, CA).

Several subclones were generated by deleting or by recloning internal restriction fragments. Sequences were primed from flanking polylinker sequence in subclones or from oligonucleotide primers synthesized to match the ends of sequence reads (primer walking). The DNA sequence was determined from multiple reads on both strands. Sequence information was compiled and analyzed through programs from DNASTAR, Inc. (Madison, WI).

DNA sequence analysis.

Gene identifications were based primarily on the results from BLAST searches (1) conducted through servers operated by the National Center for Biotechnology Information at the National Library of Medicine. Proteins with similar sequences are denoted by their locus tags or GenInfo Identifier (GI) accession numbers, as appropriate. Domains (58) are indicated by their database accession numbers: conserved domain (cd), clusters of orthologous groups (COG), or protein family (pfam).

Analysis was facilitated as unpublished whole-genome sequence data for K. pneumoniae strain MGH78578 were released by the Washington University Genome Center. These data were particularly useful for assigning likely promoters and sites for translation initiation and termination, which most likely are conserved between the two species. The K. pneumoniae MGH78578 chromosome sequence is available in GenBank under accession number NC_009648.

Allelic replacement.

Two insertion alleles were constructed directly. The hpxR113::Ω-Cm allele was made by cloning a BamHI-released Ω-Cm interposon (30) into the unique BclI site within the hpxR gene. This insertion was cloned into the conditionally replicating vector pKAS46 and transplanted to the chromosome by integration-excision allelic exchange as described previously (102). The guaD114::pAH144 allele was made by cloning an internal fragment spanning guaD codons 125 to 334 into the conditionally replicating vector pAH144. The single-crossover integration was selected by resistance to Sp. The veracity of both replacements was confirmed by PCR analysis.

Φ(hpxD-lacZ) and Φ(hpxR-lacZ) operon fusions.

The divergent hpxD-hpxR transcription control region was cloned as two approximately 320-nucleotide (nt) PCR-generated fragments with introduced XbaI and BamHI restriction endonuclease sites at either end. The endpoints lie within the codons encoding Ala-30 and Gln-37 of the hpxD and hpxR coding regions, respectively. Fragments were subcloned into BamHI- and XbaI-digested vector pVJS2354 to create Φ(hpxD-lacZ) and Φ(hpxR-lacZ) operon fusions. The resulting constructs were transplanted into the rha locus by integration-excision allelic exchange as described previously (103).

Nucleotide sequence accession number.

The DNA sequence reported in this paper is available from the GenBank nucleotide sequence database under accession number EU884423. The sequence spans 23,740 nt, including 287 nt comprising the 3′ portion of a gene homologous to the Serratia marcescens fbpC gene (2) (adjacent to the hpxE gene) and 581 nt comprising the 3′ portion of a gene homologous to Erwinia carotovora gene ECA2135 (adjacent to the hpxK gene).

RESULTS AND DISCUSSION

Purine utilization.

K. oxytoca M5al can use purines, pyrimidines, and most amino acids as sole sources of nitrogen (66). The purine rings contain four N atoms (Fig. 1), so we wished to determine if all four can be assimilated. We measured the growth yield of K. oxytoca cultured with limiting amounts of nitrogen, as described in Materials and Methods. Nitrate (one assimilable N atom) and histidine (two assimilable N atoms) (56) served as calibration controls. Results indicate that all four N atoms were assimilated from xanthine and that all five N atoms were assimilated from guanine (Table 3 and data not shown). We conclude that K. oxytoca quantitatively assimilates all N atoms from purines. An identical conclusion was drawn by de la Riva et al., who used adenine instead of guanine (24).

TABLE 3.

Growth yields from nitrogen sources

Compounda	No. of N atoms	OD420/mM N atomsb
Nitrate	1	0.199 ± 0.030
Histidine	2	0.217 ± 0.019
Xanthine	4	0.267 ± 0.053
Guanine	5	0.255 ± 0.033

^aCultures of strain VJSK009 were aerated overnight in MOPS defined medium with glucose and a range of growth yield-limiting concentrations of the indicated nitrogen source.

^bAverage final yield for at least five different cultures containing various nitrogen source concentrations. OD₄₂₀, optical density at 420 nm.

Insertion mutants.

We conducted several screens for transposon insertion mutants with specific blocks in purine utilization, as described in Materials and Methods. Mutants were tested for growth with hypoxanthine, xanthine, urate, allantoin, and allantoate (Fig. 1). Urea was not tested, since urease is encoded at a distinct, well-characterized locus (64). Four classes of mutants were identified. Mutants with the Hxn⁻ phenotype utilized neither hypoxanthine nor xanthine but did utilize urate, allantoin, and allantoate. Thus, these mutants are specifically blocked in the conversion of hypoxanthine and xanthine to urate (Fig. 1). Similarly, mutants with the Urt⁻, Aln⁻, and Alt⁻ phenotypes were specifically blocked in the utilization of urate, allantoin, and allantoate, respectively.

Our analysis of genes for utilizing allantoate (hpxFGHIJK), downstream purine catabolites (hpxWXYZ), and an associated regulator (hpxU) will be reported separately. Provisional functions for these genes are included in the deposited GenBank file.

We employed bacteriophage P1-mediated generalized transduction to backcross each insertion to the wild type in order to demonstrate linkage between the transposon and the Hpx⁻ phenotype. Genetic crosses (105) with pairs of insertions encoding different antibiotic resistance phenotypes established the order hpxD-hpxO-hpxB-hpxF (Fig. 2A). Additional crosses not depicted in Fig. 2A yielded congruent results. Subsequent molecular genetic analysis confirmed this order (see below and Fig. 2B).

FIG. 2.

Klebsiella genes for purine utilization. (A) Genetic map as determined by generalized transduction. Map distances are calculated as 100 minus percent cotransduction. The arrow points to the selected marker. (B) Physical map as determined by DNA sequence analysis. Genes are denoted by arrows according to their encoded products: gray, enzymes; open, transporters; filled, regulators. Filled circles show sites of transposon insertion, and filled squares show directed allelic replacements. Line arrows indicate predicted RpoN-dependent transcripts. The five genetic modules are denoted by the substrate upon which their products act. (C) Gene organization in K. pneumoniae MGH78578. See also Table 4. Locus tags (KPN_) delimiting the boundary of each unit are shown. Arrowheads indicate relative orientation in the genome.

As we completed the DNA sequence of the hpx locus (see below), we employed whole-colony PCR amplification and DNA sequencing to determine the exact insertion site for each of the transposons. The results revealed 11 distinct insertions (Table 2 and Fig. 2B). Thus, the insertion mutagenesis fell far short of saturating the hpx locus. We used allelic replacement methods to construct two additional null alleles as described below. Overall, analysis of mutant phenotypes in conjunction with predictions from DNA sequence analysis allows functional assignments for each of the genes as described below.

Gene product identification.

We cloned the hpx⁺ genes by complementation as described in Materials and Methods. Further subcloning and complementation analyses localized specific hpx functions to defined regions of the restriction map as described below. We determined the DNA sequence for the 22,872-nt hpx region, and concomitant computer-assisted analyses identified 23 genes (Fig. 2B).

Some functional assignments are based on homology to known proteins. In other instances, however, the proteins had not previously been characterized. In these cases, we initially attempted to identify orthologs present in genomes of other species (48). However, orthologous proteins do not necessarily perform identical functions (75); furthermore, it can be difficult to ascertain orthology (48). Therefore, we have identified what we presume to be “isofunctional homologs” in other species, based not only on overall sequence similarity and conservation of critical residues but also on genetic clustering of structural genes with those of other enzymes in the same or related pathways (i.e., synteny) (43). These assignments are described in detail below and are summarized in Table 4.

TABLE 4.

hpx genes, proposed functions, and isofunctional homologs

Gene	No. of codons	Proposed function	Isofunctional homologa
			KPN_	XCC	BTH_I	ACIAD	W168
hpxE	322	(Hypo)Xanthine hydroxylase reductase	01661	0296
hpxD	345	(Hypo)Xanthine hydroxylase	NAb	0297
hpxR	312	hpxDE transcription activator (LysR type)	01662	0298
hpxO	384	Urate hydroxylase	01663	0279c		3540
hpxP	460	Purine permease	01664				pucJ, pucK
hpxQ	166	OHCUd decarboxylase	01665	0280	2068	3537	pucLe
hpxT	108	5-Hydroxyisourate hydrolase	01666	0278	2072	3538	pucM
guaD	440	Guanine deaminase	01791
hpxC	495	Allantoin permease	01790		2065	2247	pucI
hpxS	238	Transcription regulator (GntR type)	01789		2064
hpxA	247	Allantoin racemase	01788		2066	2248
hpxB	310	Allantoinase	01787	0282	2067	3536

^aLocus tags: KPN_, K. pneumoniae MGH78578; XCC, Xanthomonas campestris ATCC 33913; BTH_I, Burkholderia thailandensis E264 (chromosome I); ACIAD, Acinetobacter sp. strain ADP1; W168, Bacillus subtilis W168 (genes).

^bNA, not annotated. The orthologous gene is present at this location but not currently annotated.

^cNonorthologous “flavin-containing monooxygenase,” presumably catalyzing the same reaction.

^dOHCU, 2-oxo-4-hydroxy-4-carboxy-5-ureidoimidazoline.

^eThe region encoding the amino-terminal domain only.

Transcriptional organization.

Transcription of most K. oxytoca genes encoding nitrogen assimilation functions is regulated by general nitrogen control (Ntr) acting through the NtrC or Nac activator proteins (7, 55, 62). The NtrC protein activates promoters dependent upon the RpoN (σ⁵⁴) form of RNA polymerase, which recognizes promoters whose consensus sequence is TGGYRYRRNNNYYGCW (where R represents purine, Y represents pyrimidine, and W represents A or T) (55). We identified six potential promoters by visual inspection, each of which is conserved in the K. pneumoniae MGH78578 sequence. These promoters are indicated in Fig. 2B as directing synthesis of the following hypothetical transcripts, shown by arrows: hpxO, hpxPQT, guaD, hpxC, hpxWXYZ, and hpxFGHIJK. Nucleotide sequence coordinates for these potential promoters are included in the deposited GenBank file. Thus, transcription of most hpx genes is likely activated directly by phospho-NtrC protein. We emphasize that we have not examined operon structure through direct experiments.

Modularity of hpx gene organization.

Genes apparently involved in each step of the overall pathway are clustered together, so the overall hpx locus appears to comprise a supraoperonic cluster (100) containing five distinct modules of hpx genes plus the guaD and hpxU genes (Fig. 2B). The gene content and order within modules is essentially identical in the K. pneumoniae MGH78578 genome (Table 4). However, in K. pneumoniae, the five modules are separated into three distinct groups: hpxEDR-hpxOPQT- ∼100 kb - hpxKJIHGF- hpxU-hpxWXYZ - ∼16 kb - hpxBASC-guaD (Fig. 2C). Note that the relative orientations between modules also differ in the two species.

For K. oxytoca M5al, both the genetic and the physical maps indicate that these modules are immediately adjacent. For example, the cotransduction frequency between the hpxF111::Tn5d(Sp) and hpxD103::MudJ alleles corresponds to a physical distance of roughly 30 kb, as estimated from the formula of Sanderson and Roth (81), approximately twice the actual distance of about 16 kb. Similarly, the calculated distance between the hpxO112::Tn10d(Tc) and hpxB105::MudJ alleles, roughly 15 kb, is approximately twice the actual distance of about 8 kb. Thus, at least for the region encompassing the hpx genes, the genomes of these two Klebsiella species appear to be organized differently.

Guanine to xanthine: GuaD guanine deaminase.

Guanine is deaminated to form xanthine (Fig. 1). The guaD gene product is most similar to E. coli K-12 guanine deaminase (60% sequence identity over 97% of its length), the product of the guaD (ygfP; cd01303) gene (61). The GuaD sequence includes a conserved nine-residue motif (P-G-X-V/I-D-X-H-T/V/I-H) also shared by members of the cyclic amidohydrolase family. This motif has been implicated in Mn²⁺ or Zn²⁺ binding in members of the family (46, 101). E. coli guanine deaminase contains Zn²⁺ (61).

We constructed a guaD disruption strain to determine the role of this gene in purine utilization. The resulting mutant failed to grow on plates with guanine as the sole nitrogen source but exhibited wild-type growth with adenine, hypoxanthine, xanthine, and urate. The guaD coding sequence is preceded by an apparent RpoN-dependent promoter and thus likely constitutes a single-gene operon (Fig. 2B).

K. oxytoca M5al also uses adenine as the sole nitrogen source (66). However, the hpx locus does not include the gene for adenine deaminase.

(Hypo)Xanthine to urate: the hpxR-hpxDE module.

Hypoxanthine is hydroxylated to form xanthine, which also is hydroxylated to form urate (Fig. 1). Xanthine dehydrogenase (purine hydroxylase), a well-studied molybdoflavoenzyme, is the previously described enzyme catalyzing the sequential oxidation of hypoxanthine and xanthine to urate (51). Structural genes for this enzyme have been characterized for Rhodobacter capsulatus (52) and Bacillus subtilis (84). However, no such genes are present in the K. oxytoca hpx locus, and none were identified in the K. pneumoniae MGH78578 genome sequence.

Three MudJ insertion mutants exhibited the Hxn⁻ phenotype, failing to grow with hypoxanthine or xanthine as the sole nitrogen source but displaying wild-type growth with urate and downstream purine catabolites. Two of these insertions disrupt the hpxE gene, whereas the third lies 20 nt upstream of the hpxD initiation codon (Table 2) and therefore blocks hpxDE operon transcription. We also constructed an hpxR insertion to determine the role of this gene in regulation (see below). This hpxR::Ω-Cm mutant also exhibited the Hxn⁻ phenotype. All four insertion mutants were complemented to the Hxn⁺ phenotype by plasmid pVJS3958, a subclone that contains only the hpxRAB region (Table 1). Therefore, the hpxD, hpxE, and hpxR genes encode functions required for conversion of (hypo)xanthine to urate, a conclusion reached independently by de la Riva et al. (24). The hpxD and hpxE genes are separated by only 13 nt and thus likely comprise an operon.

(i) HpxD and HpxE: (hypo)xanthine hydroxylase.

Sequence comparisons indicate that the hpxE and hpxD gene products comprise the reductase and oxygenase, respectively, of a two-component Rieske nonheme iron aromatic-ring-hydroxylating oxygenase system. The HpxD sequence is that of a Rieske nonheme iron oxygenase, categorized in the phthalate family of Gibson and Parales (34), group I of Nam et al. (68), and class IA of Batie et al. (5). These oxygenases are homomultimeric aromatic-ring-activating mono- and dioxygenases with broad substrate specificity. They share relatively low sequence identity and range in length from 329 to 446 aminoacyl residues (the HpxD enzyme is 345 residues). The HpxD sequence includes a Rieske-type iron-sulfur cluster (consensus sequence, C-X-H-X_15-26-C-X₂-H; pfam00355; COG2146) near the amino terminus and a likely nonheme Fe(II)-binding 2-His-1-carboxylate facial triad (consensus sequence, H-X_4-5-H-X_n-D) near the center (reviewed in references 31 and 73). The HpxD sequence shares 29% identity over a 160-residue span with that of the well-studied phthalate dioxygenase (OphA2 enzyme; GI 4128221; 443 residues) from Burkholderia cepacia DBO1 (6, 16).

The HpxE sequence (322 residues) is that of a corresponding reductase, categorized in class IA of Batie et al. (5), and includes the three defining cofactor domains: an amino-terminal FMN binding domain (consensus sequences, R-X-Y-S-L and G/S-R-G-G-S), a central NAD binding domain (consensus sequence, G-G-I-G-I-T-P; pfam00175) (in HpxE and KshB [95], the first Gly residue is replaced with Ala), and a carboxyl-terminal plant-type iron-sulfur cluster (consensus sequence, C-X₂-G-X-CG-X-C-X₆-G-X₃-HRD-X₂-L-X₅-A-X₇-C-X-S; pfam00111; COG0633) (in HpxE, L-X₅ is Q-X₆) (16, 18, 67, 95; reviewed in reference 60). The HpxE sequence shares 34% identity with the sequence of the phthalate dioxygenase reductase (OphA1 enzyme; GI 13432205; 322 residues) from B. cepacia DBO1 (6, 16) and about 40% identity with those of the vanillate O-demethylase reductases from a variety of species (e.g., the VanB enzyme from Acinetobacter sp. strain ADP1; GI 2271499; 318 residues) (86).

Purines, which resemble two fused aromatic rings, exhibit some aromatic properties (87). We suggest that the HpxE and HpxD enzymes comprise (hypo)xanthine hydroxylase reductase and bifunctional (hypo)xanthine hydroxylase, respectively. This enzyme complex likely functions by the same mechanisms as those described for homologous aromatic-ring-hydroxylating oxygenases, hydroxylating first the C-2 position of hypoxanthine to yield the enol form of xanthine and second the C-8 position to yield the enol form of urate (Fig. 1). Purines undergo tautomeric interconversions to shift between the relatively unstable enol form and the more stable keto form (8, 23, 87).

(ii) Transcriptional regulation in the hpxR-hpxDE module.

The two hpxE::MudJ insertion mutants expressed LacZ activity, indicating that they form Φ(hpxE-lacZ) operon fusions. We cultured these strains in defined medium, with glutamine as a neutral nitrogen source, and measured LacZ-specific activity. Both strains expressed approximately 4,000 Miller units after growth with added xanthine but less than 50 units after growth with added urate, allantoin, or allantoate and about 10 units after growth with added xanthine plus ammonium. This indicates that hpxDE operon transcription is subject to dual control by global nitrogen regulation (Ntr) and by pathway-specific induction by (hypo)xanthine (see also below) but not downstream purine catabolites.

The hpxDE operon shares a 125-nt intergenic region with the divergently transcribed hpxR gene (Fig. 2B). The HpxR sequence (312 residues) is similar to those of LysR family transcriptional regulators, many of which regulate degradation pathways for aromatic compounds (93). The HpxR sequence includes an amino-terminal helix-turn-helix motif (pfam00126) and a carboxyl-terminal LysR-type substrate-binding domain (pfam03466).

We constructed an hpxR::Ω-Cm insertion strain to determine the role of this gene in purine utilization. We also constructed monocopy Φ(hpxD-lacZ) and Φ(hpxR-lacZ) operon fusions at a separate locus in order to evaluate expression in hpx⁺ strains. We cultured both hpxR⁺ and hpxR::Ω-Cm derivatives of these strains in defined medium, with nitrate as a neutral nitrogen source, and measured LacZ-specific activity. The results are shown in Table 5. Expression from the Φ(hpxD-lacZ) hpxR⁺ strain was subject to both Ntr and pathway-specific control, consistent with the results from the hpxE::MudJ strains summarized above. In contrast, expression from the Φ(hpxD-lacZ) hpxR::Ω-Cm strain was undetectable, demonstrating that the HpxR protein is essential for activating hpxDE operon transcription. This is consistent with the Hxn⁻ phenotype conferred by the hpxR::Ω-Cm insertion noted above. Expression from the Φ(hpxR-lacZ) strain was essentially indifferent to added ammonium or hypoxanthine but was elevated about threefold in the hpxR::Ω-Cm insertion strain relative to the level for the hpxR⁺ strain. This suggests that hpxR transcription is subject to weak negative autoregulation.

TABLE 5.

Effects of hypoxanthine and ammonium on expression from Φ(hpxD-lacZ) and Φ(hpxR-lacZ) constructs

Strain	Fusionb	hpxRc	LacZ sp acta
			− Hyx/ + NH4+	− Hyx/ − NH4+	+ Hyx/ − NH4+
VJSK2966	Φ(hpxD-lacZ)	+	7	42	2,420
VJSK2968	Φ(hpxD-lacZ)	−	<1	<1	<1
VJSK2967	Φ(hpxR-lacZ)	+	30	40	29
VJSK2969	Φ(hpxR-lacZ)	−	100	200	180

^aStrains were cultured to the mid-exponential phase in MOPS medium, with nitrate as the nitrogen source. Ammonium or hypoxanthine (Hyx) was added as indicated.

^bOperon fusion construct integrated at the chromosomal rha locus.

^c+, hpxR⁺; −, hpxR113::Ω-Cm.

These results indicate that the HpxR protein is activated by binding (hypo)xanthine to stimulate hpxDE operon transcription initiation. The source of Ntr regulation is less clear. No RpoN-dependent promoter is evident in the hpxD-hpxR intergenic region in either K. oxytoca M5al or K. pneumoniae MGH78578. Expression of the Φ(hpxR-lacZ) fusion was not regulated by Ntr, indicating that Ntr regulation does not act by controlling HpxR protein synthesis.

Very similar results were reported by de la Riva et al. (24), who further demonstrated specific binding of HpxR protein to the intergenic regulatory region. They additionally concluded that Ntr control is mediated by neither the NtrC nor the Nac proteins (7).

(iii) HpxD, HpxE, and HpxR isofunctional homologs in other species.

The HpxD and HpxE sequences are most closely related to sequences from Xanthomonas spp. (63% and 47% identity, respectively), currently annotated as vanillate O-demethylase. Likewise, the HpxR sequence is most closely related to sequences from Xanthomonas spp. (46% identity). The genes encoding these proteins are adjacent to several other genes involved in purine catabolism (Table 4). Therefore, these genes comprise the hpxR-hpxDE module in Xanthomonas spp.

Urate to S-(+)-allantoin: the hpxO-hpxPQT module.

Urate is hydroxylated to form 5-hydroxyisourate (HIU), which is hydrolyzed to form 2-oxo-4-hydroxy-4-carboxy-5-ureidoimidazoline (OHCU), which is decarboxylated to form S-(+)-allantoin (92) (Fig. 1). HIU also undergoes spontaneous conversion to racemic allantoin (45). These reactions comprise the first ring-opening step in purine utilization.

Uricase (urate oxidase) is the previously described enzyme catalyzing the initial oxidation of urate to HIU. Structural genes for this enzyme have been characterized for Bacillus subtilis (84) and for a range of eukaryotic species (71). However, no such genes are present in the K. oxytoca hpx locus, and none were identified in the K. pneumoniae MGH78578 genome sequence.

Two insertion mutants exhibited the Urt⁻ phenotype, failing to grow with hypoxanthine, xanthine, or urate as the sole nitrogen source but displaying wild-type growth with allantoin and allantoate. Both insertions disrupt the hpxO gene (Table 2). Therefore, the hpxO gene encodes a function required for utilization of urate, a conclusion reached independently by de la Riva et al. (24).

The hpxP and hpxQ genes overlap (23 nt), as do the hpxQ and hpxT genes (4 nt), and thus likely comprise the hpxPQT operon (Fig. 2B). Apparent RpoN-dependent promoters for both the hpxO gene and the divergently transcribed hpxPQT operon are present in the 346-nt intergenic region.

(i) HpxO: urate hydroxylase.

Sequence comparisons classify the hpxO gene product (42 kDa) in the ≈45-kDa aromatic-ring flavoprotein monooxygenase family (pfam01360), members of which share relatively low sequence similarity (40). For example, the HpxO sequence shares 28% sequence identity over 84% of its length with salicylate hydroxylase (NahG enzyme; GI 4104764; 437 residues) from Pseudomonas stutzeri AN10 (9).

Enzymes in this family cleave dioxygen without the use of a metal ion and incorporate one oxygen atom as a hydroxyl group ortho or para to an existing hydroxyl group while reducing the other oxygen atom to water (33; reviewed in reference 40). The HpxO sequence (384 residues) comprises three defining cofactor domains: an amino-terminal domain that binds the ADP moiety of flavin adenine dinucleotide (FAD) (consensus sequence, G-X-G-X₂-G-X₃-A/G-X₆-G; pfam01494), a central domain that binds the pyrophosphate moiety of NAD(P)H and interacts indirectly with the pyrophosphate moiety of FAD (consensus sequence, D-X₅-DG-X₅-R; pfam01360), and a carboxyl-terminal domain that binds the ribose moiety of FAD. Sequences in the last domain vary considerably, save for a conserved G-D sequence (10, 28, 29, 59; reviewed in reference 40).

p-Hydroxybenzoate hydroxylase (PobA enzyme) is the archetype for the ≈45-kDa aromatic-ring flavoprotein monooxygenase family (3). The reaction mechanism initiates with binding of the substrate (p-hydroxybenzoate). This leads to the reduction of the flavin isoalloxazine ring by NADPH, forming reduced FAD. Dioxygen then binds to the complex and is reduced by two single-electron transfers. The resulting C4a-flavin (hydro)peroxide hydroxylates the substrate (33; reviewed in reference 40). The HpxO enzyme likely uses a similar mechanism to hydroxylate urate to form HIU (Fig. 1).

(ii) HpxT and HpxQ.

Sequence comparisons indicate that the HpxT and HpxQ proteins are HIU hydrolase and OHCU decarboxylase, respectively (15, 44, 77). The HpxT sequence (108 residues; COG2351) is 32% identical to that of HIU hydrolase (PucM protein; 121 residues) from Bacillus subtilis 168 (44, 84). The HpxQ sequence (166 residues; COG3195) is 27% identical to that of the amino-terminal 165 residue OHCU decarboxylase domains of urate oxidase (uricase) from Bacillus sp. strain TB-90 (GI 21431959) (104) and B. subtilis 168 (PucL protein) (84). The Bacillus uricase is synthesized as a precursor of about 500 residues, from which the amino-terminal OHCU decarboxylase is proteolytically cleaved (70).

(iii) HpxP purine permease.

The HpxP sequence (460 residues) shares 32% identity over 95% of its length with the high-affinity, high-capacity urate-xanthine permease UapA (GI 6136091; 615 residues) (36) and 30% identity over 90% of its length with the general purine permease UapC (GI 1351342; 580 residues), both from A. nidulans (26). Additionally, the HpxP sequence is 34% identical to those of the urate permeases PucJ (449 residues) and PucK (430 residues) from B. subtilis 168. The HpxP sequence meets the criteria for inclusion in the nucleobase ascorbate transporter family (COG2233) (37), including the nucleobase ascorbate transporter signature motif (Q/E/P-N-X-G-X₄-T-R/K/G), the QH motif in the middle of transmembrane segment 1, and a conserved alanyl residue important for function but not specificity (Ala-404 in UapA and Ala-272 in HpxP).

The region spanning residues 378 to 446 in UapA and 336 to 404 in UapC (analogous to the region spanning residues 246 to 314 in HpxP) determines uptake specificity, with UapA residues Glu-412 and Arg-414 and UapC residues Gln-370 and Glu-372 possibly forming part of a purine binding site (27). The corresponding residues in the HpxP sequence are more similar to those of UapC than to those of UapA: Ser-280_HpxP and Gln-370_UapC are both uncharged, with polar side chains, whereas Glu-412_UapA is acidic, and Asp-282_HpxP and Glu-372_UapC are both acidic, whereas Arg-414_UapA is basic. This suggests that HpxP may be a general purine permease, an appealing hypothesis since this is the only hpx cluster-encoded permease for transport of guanine, hypoxanthine, xanthine, and urate.

(iv) HpxO isofunctional homologs in other species.

The HpxO sequence is most similar (53% identical) to a protein sequence annotated as “putative flavoprotein monooxygenase acting on aromatic compound” from Acinetobacter sp. strain ADP1 (locus tag ACIAD3540; GI 49532471; 385 residues). This gene is adjacent to those encoding HIU hydrolase and OHCU decarboxylase (Table 4), supporting its assignment as encoding the isofunctional homolog of HpxO protein. Intriguingly, Xanthomonas genomes contain a gene for a presumptive flavin-containing monooxygenase (e.g., XCC0279), not homologous to HpxO, adjacent to those encoding HIU hydrolase and OHCU decarboxylase (Table 4). Thus, it appears that at least two different monooxygenases have been recruited into the urate hydroxylase step of the purine utilization pathway.

R-(−)-allantoin to S-(+)-allantoin to allantoate: the hpxC-hpxSAB module.

Conversion of allantoin to allantoate is the second of the two ring-opening steps in purine utilization (Fig. 1). Five insertion mutants exhibited the Aln⁻ phenotype, failing to grow with allantoin or upstream purine catabolites as the sole nitrogen source but displaying wild-type growth with allantoate. The insertions disrupt the hpxA and hpxB genes (Table 2). Therefore, the hpxAB operon encodes functions involved in allantoin utilization. However, the hpxA::Tn10d(Tc) insertion was complemented by subclones carrying the hpxB but not the hpxA gene, indicating that the hpxA gene is not essential for growth on allantoin. Presumably, the hpxA::Tn10d(Tc) insertion is polar on hpxB gene expression.

Allantoinase is the previously described enzyme catalyzing the conversion of allantoin to allantoate (91). Structural genes for this enzyme have been characterized for Bacillus subtilis (84) and E. coli (21) and for a range of eukaryotic species (12). However, no such genes are present in the K. oxytoca hpx locus, and none were identified in the K. pneumoniae MGH78578 genome sequence.

Most characterized allantoinases are specific for the S isomer. However, the nonenzymatic conversion of HIU to allantoin creates a racemic mixture of R- and S-allantoin (45), and allantoin is present as the racemic mixture in decaying organic matter (94). Allantoin racemase has been characterized for bacteria (94) and for Candida utilis (72). To our knowledge, the structural gene for an allantoin racemease has not been identified.

An apparent RpoN-dependent promoter for the hpxC gene (but not for the hpxSAB operon) is present in the 52-nt hpxC and hpxSAB intergenic region. The hpxS and hpxA genes overlap (1 nt; TAATG), and the hpxA and hpxB genes are separated by only 16 nt, so these likely comprise the hpxSAB operon (Fig. 2B).

(i) HpxB: allantoinase.

Sequence comparisons suggested that the HpxB enzyme catalyzes conversion of allantoin to allantoate. The HpxB sequence (310 residues; COG0726) shares 26% identity over 74% of its length with an imidase (291 residues; GI 14040045) from Ralstonia eutropha (97). The imidase catalyzes the hydrolysis of hydantoin and related compounds. Since allantoin is a 5-substituted hydantoin (ureidohydantoin), this suggests that the HpxB enzyme acts similarly to the imidase by hydrolyzing a cyclic C-N bond in the hydantoin ring to form allantoate (Fig. 1).

This conjecture was confirmed by Ramazzina et al., who recently identified the HpxB homolog from Pseudomonas fluorescens DSM 50090 as an (S)-allantoin-specific allantoinase (76). They denote this enzyme PuuE.

(ii) HpxA: allantoin racemase.

Sequence comparisons indicate that the HpxA protein is allantoin racemase. The HpxA sequence (247 residues; COG 4126; pfam01177) shares 45% identity over 96% of its length with 5-substituted hydantoin racemase (HyuE enzyme; 229 residues; GI 266318) from Pseudomonas sp. strain NS671 (98). The HyuE enzyme catalyzes racemization of a variety of 5-substituted hydantoins (99). Since HpxB allantoinase is specific for the S enantiomer (76), allantoin racemase could be important for efficient allantoin utilization in natural environments. However, allantoin racemase likely is not required for colony formation on plates with abundant (racemic) allantoin, explaining why the hpxB⁺ gene was able to complement the polar hpxA insertion (see above).

(iii) HpxC: allantoin permease.

The HpxC sequence (495 residues; COG1953) shares 22% identity over 85% of its length with allantoin permease from Saccharomyces cerevisiae (Dal4p protein; 635 residues) and 33% identity over 92% of its length with allantoin permease from Bacillus subtilis W168 (PucI protein; 490 residues).

(iv) HpxS.

Sequence comparisons indicate that the HpxS protein is a member of the GntR superfamily of transcriptional regulators (COG1802). For example, the HpxS sequence (238 residues) shares 28% identity over 58% of its length with the MatR repressor (222 residues; GI 6840861), which controls expression of the matABC operon responsible for the uptake and conversion of malonate to acetyl-coenzyme A in Rhizobium leguminosarum (50).

The GntR family is made up of bacterial transcriptional regulators with similar amino-terminal helix-turn-helix DNA-binding domains and divergent carboxyl-terminal effector-binding and oligomerization domains (93). The GntR superfamily has been subdivided into four subfamilies on the basis of secondary structure in the C-terminal domain (78). On the basis of these considerations, the HpxS protein can be classified into the VanR subgroup of the FadR subfamily (COG2186). Many proteins in the FadR subfamily regulate operons involved in amino acid metabolism. Allantoin resembles a dipeptide, and the bond cleaved to yield allantoate resembles a peptide bond (Fig. 1).

(v) Transcriptional regulation in the hpxC-hpxSAB module.

The hpxB105::MudJ mutant expressed LacZ activity, indicating that this insertion forms a Φ(hpxB-lacZ) operon fusion. We cultured this strain in defined medium, with glutamine as a neutral nitrogen source. LacZ-specific activity was approximately 1,600 Miller units after growth with added xanthine and about 10 units after growth with added xanthine plus ammonium. This indicates that hpxSAB operon transcription is subject to global nitrogen regulation (Ntr). LacZ-specific activity was about 1,050 units after growth with added allantoin, about 775 units with urate, and about 400 to 500 units with allantoate or with no addition, suggesting possible pathway-specific induction by allantoin or an upstream intermediate. Presumably, the HpxS protein controls pathway-specific transcriptional regulation of the hpxSAB operon.

(vi) HpxA and HpxB isofunctional homologs in other species.

The HpxB allantoinase sequence is most similar (about 70% identity) to protein sequences typically annotated as “polysaccharide deacetylase” (76), and the HpxA sequence is most similar (up to 62% identity) to protein sequences typically annotated as “hydantoin racemase.” Representative examples are shown in Table 4. In the case of Burkholderia spp. and related species, the hpxA and hpxB homologs are present as an hpxSCAB module within a larger cluster of purine utilization genes, including those for HIU hydrolase and OHCU decarboxylase. In contrast, the Acinetobacter sp. hpxB homolog is clustered with the hpxO, hpxT, and hpxQ homologs, encoding enzymes of urate catabolism, whereas the hpxA and hpxC homologs are adjacent to each other at a separate location. This indicates that the HpxA allantoin racemase may function physiologically to scavenge exogenous, racemic allantoin. Consistent with this notion, the Xanthomonas spp. have hpxB homologs but none for hpxA or hpxC.

Van der Drift et al. identified allantoin racemase in P. fluorescens and Pseudomonas putida (among others) but not in Pseudomonas aeruginosa or P. stutzeri (94). Database searches reveal that hpxA homologs are present in genomes of P. fluorescens, P. putida, and Pseudomonas syringae but not in P. aeruginosa or P. stutzeri (locus tag Pput_1557 in P. putida F1). This reinforces the conclusion that hpxA encodes allantoin racemase. In these genomes, the hpxA and hpxC genes are adjacent to each other but not to other purine utilization genes, as described above for the Acinetobacter sp. (The P. syringae clusters include an hpxS homolog as well.) In contrast, the hpxB gene, encoding allantoinase, is present in the genomes of all five Pseudomonas species examined (locus tag Pput_1582 in P. putida F1), clustered with several other genes involved in purine utilization.

The HpxA sequence shares 25% identity with the Dcg1p proteins of S. cerevisiae (244 residues; locus tag YIR030C) and Candida albicans (232 residues; locus tag CaO19.244). The S. cerevisiae DCG1 gene was identified as a nitrogen-regulated gene within the DAL cluster, encoding allantoinase, allantoin permease, and allantoicase (106), but the function of the Dcg1p protein has not been determined. Allantoin racemase activity has been characterized for Candida utilis (72). The analysis presented here indicates that Dcg1p may function as allantoin racemase.

Conclusions.

Microbial genome sequencing projects reveal large numbers of genes with unknown functions. Here, Klebsiella purine utilization (Fig. 1) provides numerous examples of unpredicted enzymes catalyzing an established pathway (96).

Several enzymes were evidently recruited from pathways for catabolism of aromatic compounds. Aromatic compounds and the pathways for their catabolism are diverse and widespread, and so, in retrospect, it is unsurprising that such enzymes are involved in oxidizing purines (Fig. 1).

Oxidation of (hypo)xanthine to form urate is catalyzed by a two-component Rieske nonheme iron aromatic-ring-hydroxylating oxygenase system (HpxE-HpxD enzyme), and oxidation of urate to form 5-hydroxyisourate is catalyzed by an ≈45-kDa aromatic-ring flavoprotein monooxygenase (HpxO enzyme). These enzymes were identified independently by de la Riva et al. (24). Genome sequence comparisons suggest that a distinct flavin-containing monooxygenase catalyzes the latter step in Xanthomonas spp. (Table 4).

Yet another enzyme is involved in xanthine utilization by A. nidulans, for which purine utilization has long been studied (83). Early work established that mutants lacking xanthine dehydrogenase (hxA) or molybdenum cofactor (cnx) exhibit leaky growth with xanthine (but not hypoxanthine) as the sole nitrogen source (22). This indicated the presence of a molybdenum cofactor-independent xanthine alternative pathway, dependent on the xanA gene (85). Recently, it was established that XanA protein, a member of the taurine dioxygenase group of enzymes active on aromatic compounds, is an α-ketoglutarate-dependent dioxygenase that hydroxylates xanthine to form urate (20, 65).

Finally, our analysis reveals that Klebsiella spp. employ HpxA allantoin racemase, related to hydantoin racemases (72, 94), as well as the HpxB (PuuE) allantoinase, related to polysaccharide deacetylases and described recently by Ramazzina and coworkers (76). Our work in progress indicates further that other previously undescribed enzymes are involved in allantoate catabolism (Fig. 2).