The Rut Pathway for Pyrimidine Degradation: Novel Chemistry and Toxicity Problems

Bacterial strains.The three strain backgrounds we worked in were wild type (NCM3722), ntrB(Con) (ntrB [constitutive]; NCM3876), and UpBCon1 (NCM4384). Strains carrying in-frame deletions in each rut gene and ydfG were constructed in each background by first introducing the appropriate kanamycin (Kan) insertion by bacteriophage P1-mediated transduction and then deleting it by site-specific recombination (11) (Table 1). The correctness of all deletions was confirmed by sequencing. Due to the absence of four bases in the forward primer for rutD (4), RutC extends an additional 36 amino acids in the initial rutD deletion strains (Table 1). We do not know how this affects RutC activity. The inadvertent change to RutC was corrected by cloning the rutC rutD::Kan fragment from NCM4075 (derived by P1-mediated transduction from NCM4053; Table 1) into the pGEM-T Easy vector and correcting the sequence of the forward primer by site-directed mutagenesis (from ATATCGCAAGTGGGCGCGAGATTCCGGGATC [incorrect] to ATATCGCGAAGTGAGGCCGCGATGATTCCGGGATC [correct]). After the change to rutC was corrected, rutD::Kan was introduced into a ΔrutC strain (Table 1) (11) and used to generate correct rutD deletions in different backgrounds. The rutF deletion extended an additional 12 bp beyond its predicted C-terminal end but remained in-frame and within rutF. Since ydfG is the only gene in its operon, it was not necessary to delete it unless kanamycin sensitivity was required.

Identifications of mutations in strains NCM4139, NCM4299, NCM4300, and NCM4384.An 11-kbp deletion beginning in the mioC gene was first found in strain NCM4384 based on tiling microarray data from Roche Nimblegen (23). The extent of the deletion was verified by PCR amplification and sequencing, and the same deletion was found to have occurred independently when a rutE::Kan mutation was introduced into the ntrB(Con) background by phage P1-mediated transduction (Table 1). Roche 454 deep sequencing (20- to 30-fold coverage) allowed us to identify the remaining mutations. Using E. coli K-12 strain MG1655 as a reference strain for the assembly of each of the four genomes, Roche provided tables of differences between each strain and MG1655. These tables were used as a starting point to find mutational changes. By manual inspection of raw sequence data, sequence differences between strains, and contig breaks, we found independent single-base-pair changes associated with nemR in strains NCM4139, NCM4299, and NCM4300 and sroG in strain NCM4384 (see Results). We then identified the IS186 insert in the lon promoter in strains NCM4139 and NCM4384 by looking for new occurrences of the small insertion elements (IS1, IS2, IS3, IS4, IS5, IS30, IS150, and IS186).

Growth and toxicity studies.Growth studies were done in N⁻ C⁻ minimal medium containing uridine or thymidine as the sole nitrogen source with 0.4% glycerol as the carbon source (31). Studies on solid medium were done with 5 mM uridine or thymidine, as were standard studies on liquid medium. Unless stated otherwise, they were done at room temperature (≤22°C). To measure toxicity, uridine or thymidine (5 mM) was added to ammonium (5 mM) as a nitrogen source and growth was monitored at 37°C.

Cell extracts of UpBCon1 (NCM4384).Cells were grown on minimal medium with glycerol (0.5%) as a carbon source and uridine (5 mM) as the sole nitrogen source at 37°C. They were harvested, washed in 20 mM potassium phosphate buffer (pH 7), and frozen at −80°C. Cells (∼0.1 g wet weight/ml) were suspended in potassium phosphate buffer (pH 7) and were disrupted in a French pressure cell (SLM Aminco Instruments, Inc., Rochester, NY) at 6,000 lb/in².

Partial purification of His-tagged proteins.ASKA strains JW0997, JW5138, and JW1532, which overproduce His-tagged RutA, RutF, and YdfG protein, respectively (26), were grown on Luria broth containing chloramphenicol (25 μg/ml) to an optical density at 600 nm (OD₆₀₀) of ∼0.5, and expression was induced for 3 to 4 h by addition of IPTG (isopropyl-β-d-thiogalactopyranoside) (to 100 μM). Cells were harvested and frozen at −80°C until use. They were suspended at ∼0.1 g wet weight/ml in 20 mM phosphate buffer (pH 7) and were disrupted as described above. His-tagged RutA protein was then partially purified using Ni⁺-nitrilotriacetic acid (NTA) agarose (Qiagen, Valencia, CA) according to the manufacturer's instructions and was finally dialyzed four times against 500 ml of 20 mM phosphate buffer (pH 7) at 4°C and stored in small aliquots at −80°C (see Fig. S1 in the supplemental material). The RutF protein was not well overexpressed and was assayed only in extracts. Protein concentrations were determined using the Micro bicinchoninic acid (BCA) protein assay kit (Thermo Scientific, IL). The Fre flavin reductase, a kind gift from Luying Xun, Washington State University, Pullman, was purified as described previously (58) and was stored at −80°C. Before use, it was diluted in 20 mM phosphate buffer (pH 7) containing 1 mM dithiothreitol (DTT). The RutB protein was used from extracts of JW5139. His-tagged RutB protein was a kind gift from Tathagata Mukherjee and Tadhg Begley, Cornell University and Texas A and M University, respectively.

Materials for the RutA/F reaction.[¹⁴C-2]- and [¹⁴C-6]uracil were purchased from MP Biomedicals (Solon, OH). [¹⁴C-methyl]thymine was purchased from Moravek Biochemicals and Radiochemicals (Brea, CA). Uniformly labeled [¹³C, ¹⁵N]uracil (99%, 98%), [¹⁵N]uracil (98%), [¹³C-4, ¹³C-5]uracil (99%), and ¹⁸O₂ (97%) were purchased from Cambridge Isotope Laboratories, Inc. (Andover, MA).

Assay for the RutA protein.Reaction mixtures contained 40 mM Tris buffer (pH 8.2) or 40 mM phosphate buffer (pH 7), 20 μM flavin mononucleotide (FMN), 4 mM NADH, and 0.4 mM uracil. For standard reactions, we used 18 μg of His₆-RutA, 6 μg of Fre, and [¹⁴C]uracil (radiolabeled at position 2 or 6 at 2 × 10⁷ cpm) in a total volume of 120 μl. Reaction mixtures were incubated at room temperature with agitation for 20 min, and reactions were stopped by putting them on ice and then freezing them at −20°C. Products were analyzed on cellulose F thin-layer chromatography (TLC) plates (Merck, Germany) in developing solution containing isopropyl alcohol-water (3:1).

For ¹⁸O₂ labeling of the RutA product, the total volume of reaction mixtures was increased 5-fold. All components except enzymes were mixed, and reaction mixtures were bubbled with N₂ for 5 min. They were then bubbled with ¹⁸O₂ or ¹⁶O₂ for 1 min. During bubbling with O₂, His₆-RutA and Fre were added and bubbling was continued for an additional 5 min. For analysis of 50:50 mixtures of ¹⁸O₂- and ¹⁶O₂-labeled products, equal volumes of separate reaction mixtures were combined.

Synthesis of Z-3-ureido-2-propenoic acid. Z-3-Ureido-2-propenoic acid (ureidoacrylate) was synthesized by adding 3 ml of 4 M ammonium hydroxide to 100 mg of 2,3-dihydro-1,3-6H-oxazine-2,6-dione (3-oxauracil) (Research Organics, Inc., Cleveland, OH) on ice as described previously (16). The reaction was allowed to proceed for 12 h at room temperature, after which 1 ml of 1 M NaOH was added, and the solution was lyophilized to dryness. The product was extracted by adding 1 ml of methanol to the dried powder. The product was obtained by lyophilizing the methanol fraction. The presence of the correct product was confirmed by comparing ¹H-nuclear magnetic resonance (NMR) chemical shifts and coupling constants in dimethyl sulfoxide (DMSO) with published values (16).

1D ¹³C NMR spectra.For NMR studies, the RutA reaction was performed using uniformly ¹³C/¹⁵N-labeled uracil (Cambridge Isotope Labs, Andover, MA) as the substrate, unless stated otherwise. Reaction mixtures prepared and frozen as described above were used as such or were lyophilized and dissolved in DMSO. One-dimensional (1D) ¹³C spectra were recorded on a Bruker Avance 600-MHz spectrometer equipped with a CPTXI cryoprobe in 4 to 16 h and a Bruker Avance 800-MHz spectrometer equipped with a TXI probe in 24 to 48 h. In all cases, ¹H decoupling was applied during acquisition, and data were zero-filled once before analysis. Spectra recorded in H₂O were referenced to 4,4-dimethyl-4-silapentane-1 sulfonic acid (DSS; 0 ppm), and samples dissolved in DMSO were referenced to tetramethylsilane (TMS; which replaces the DMSO resonance at 39.5 ppm). For complete ¹³C spectra of the product, the carrier and spectral width were set to 127 ppm and 130 ppm, respectively, and the final digital resolution was approximately 6 Hz/point. For identification of ¹⁶O-to-¹⁸O isotope shifts, RutA reactions were performed with either ¹⁶O₂ or ¹⁸O₂ (Cambridge Isotope Labs, Andover, MA) and [¹³C-4, C-5]uracil (Cambridge Isotope Labs, Andover, MA); equal amounts of the two reaction mixtures were combined. Spectra were taken in DMSO, and data were recorded at 800 MHz. The ¹³C carrier frequency and spectral width were set to 164 ppm and 25 ppm, respectively. The final digital resolution was 1.2 Hz/point.

2D NMR spectra.A 2D ¹H-¹³C heteronuclear single quantum coherence (HSQC) spectrum (43) of the product was recorded at 800 MHz in D₂O by lyophilizing the sample from H₂O and redissolving it in 100% D₂O. A total of 1,024 and 512 points were collected in the ¹H and ¹³C dimensions, respectively. The carrier frequencies were set to 5.2 ppm (¹H) and 118 ppm (¹³C), and the spectral widths were set to 13 ppm (¹H) and 80 ppm (¹³C). The spectrum was recorded in 18 h and was processed with NMRPipe software (12). After zero-filling twice in the ¹³C dimension, the digital resolution was 16 Hz/point. Chemical shifts were indirectly referenced to DSS (57).

The number of protons attached to the product N-3 resonance was determined by examining the ¹H-¹⁵N coupling pattern in a 2D ¹³C-detected ¹⁵N-¹³C HSQC experiment (6) on the reaction mixture in H₂O. The spectrum was recorded on an Avance II 900-MHz instrument equipped with a CPTXI cryoprobe using the C_CO_N HSQC pulse sequence supplied by Bruker-Biospin, Inc. However, no ¹H decoupling was applied in the ¹⁵N dimension, and no ¹⁵N decoupling was applied during detection of ¹³C. Totals of 8,192 points and 120 points were collected in the ¹³C and ¹⁵N dimensions, respectively. The carrier frequencies were set to 170 ppm (¹³C) and 95 ppm (¹⁵N), and spectral widths were set to 80 ppm (¹³C) and 87.5 ppm (¹⁵N). The total experiment time was 14 h. The data were processed with NMRPipe software (12). Data in the ¹⁵N dimension were increased to 256 points by linear prediction and subsequently to 512 points by zero-filling. The final resolutions were 4.4 Hz and 15.6 Hz in the ¹³C and ¹⁵N dimensions, respectively.

MS.Liquid chromatography-mass spectrometry (LC/MS) data were obtained using an Agilent 1200 liquid chromatograph coupled to an LTQ Orbitrap mass spectrometer. The Orbitrap mass spectrometer was operated in positive-mode electrospray ionization with a mass resolution of 30,000. A Phenomenex Capcell C₁₈ column (5μ, 120 Å, 150 by 4.6 mm) with a flow rate of 0.2 ml/min was used for optimum chromatographic separation. For the ureidoacrylate compounds, an isocratic gradient of 10% acetonitrile, 89% water, and 1% formic acid was used. For the peroxy form, a gradient of 0 to 50% methanol over 20 min was used.

Assay for the RutB protein.The RutB protein was assayed using ¹⁴C-labeled RutA product as the substrate and monitoring by TLC as described above or using chemically synthesized ureidoacrylate (16) and monitoring the decrease in absorbance at 266 nm with extinction coefficients of 17,800 M⁻ Cm⁻ in 0.025 M HCl (16) and 18,300 in 40 mM Tris buffer (pH 8.2). Formation of ammonia was monitored by coupling to the glutamate dehydrogenase reaction and measuring NADPH oxidation (ammonia assay kit; Sigma, St. Louis, MO). Formation of malonic semialdehyde was assayed by coupling to YdfG (18). Reaction mixtures contained 40 mM Tris (pH 8.2), 0.25 mM ureidoacrylate, 0.8 mM NADPH, 28 μg of RutB, and 4.5 μg of YdfG in a total volume of 400 μl. They were incubated at room temperature for 3 h. Consumption of ureidoacrylate and of NADPH was determined using their extinction coefficients, ammonium was determined as described above, and 3-hydroxypropionic acid was identified and quantified by GC/MS as described previously (31). We verified that YdfG oxidized serine and 3-hydroxypropionic acid as described previously (18). Reaction mixtures contained 200 mM Tris (pH 8.5), 1 mM NADP, 12 μg YdfG, and 0.5 M l-serine or 3-hydroxypropionic acid in a total volume of 400 μl. NADPH formation was monitored at 340 nm at room temperature. The 3-hydoxypropionic acid (NH₄⁺ salt; 138 mg/ml) was kindly provided by Hans Liao (Cargill Corporation, Minneapolis, MN).

Isolation of strains that utilize pyrimidines as the sole nitrogen source at 37°C.In a wild-type E. coli K-12 strain, the rut operon allows growth on pyrimidines as the sole nitrogen source at temperatures up to ∼22°C but not higher (31). Expression of the operon is elevated in a strain that carries an ntrB(Con) (ntrB [constitutive]) mutation, which increases transcription of all genes under NtrC control (59). In an ntrB(Con) background (NCM3876), we isolated two strains that could grow on uridine at 37°C, UpBCon1 (NCM4384) and UpBCon2 (NCM4139). The former, which grew fastest, excreted a yellow compound into the medium and grew poorly at room temperature even on enriched medium. At 37°C, it obtained both nitrogens from uridine in usable form and excreted 1 mol/mol of 3-hydroxypropionic acid into the medium (see Table S1 in the supplemental material) (31).

Degradation of ¹⁴C-labeled uracil or thymine by cell extracts.To initiate studies of Rut enzymes in vitro, we first prepared cell extracts of NCM4384 (UpBCon1) grown on uridine at 37°C. Bioinformatic predictions were that the RutA protein was a monooxygenase with alkane sulfonate monooxygenase as its closest homologue and that the RutF protein was a flavin reductase with the HpaC protein, which functions in the oxidation of 4-hydroxyphenylacetate in E. coli W, as its closest homologue (13, 19, 31). In the presence of cell extract, FMN, and NADH, [¹⁴C]uracil labeled at C-6 or C-2 was consumed (Fig. 2 A) (data not shown). FAD worked much less well than FMN, and NADPH worked much less well than NADH. HpaC also has a strong preference for FMN and NADH (19). As the amount of extract was increased, only a small amount of label from C-6 of the uracil ring remained on cellulose TLC plates, near the origin (Fig. 2A), and label from C-2 was completely lost.

• Open in new tab
• Download powerpoint

FIG. 2.

RutA acts on [¹⁴C]uracil and thymine. (A) Products from [¹⁴C-6]uracil in the presence of increasing amounts of cell extract (μl) from NCM4384 (UpBCon1) grown on uridine at 37°C. (B) Products from [¹⁴C-6]uracil (lanes 1 to 3) or [¹⁴C-2]uracil (lanes 4 to 6) in the presence of RutA (lanes 2 and 5) or RutA and RutB (lanes 3 and 6). Samples in lanes 1 and 4 are from controls with no enzyme. (C) Products from [¹⁴C-6]uracil (lanes 4 and 5) or [¹⁴CH₃]thymine (lanes 6 and 7) in the presence of RutA. Samples in lanes 1 and 2 are from controls with no enzyme, and the sample in lane 3 contained a mixture of unlabeled thymine (T), uracil (U), and barbituric acid (B), which were detected by UV absorbance and have been circled. All reactions mixtures that contained RutA also contained the flavin reductase Fre. Reactions were run for 20 min at room temperature with agitation as described in Materials and Methods, and the mixtures were frozen at −20°C before being analyzed by thin-layer chromatography. For samples in panels A and C, reactions were run at pH 8.2, and for samples in panel B, they were run at pH 7.

The RutA/F reaction.His-tagged RutA was well overexpressed from ASKA strain JW0997 and could be purified by Ni²⁺ chelate chromatography (Materials and Methods) (see Fig. S1 in the supplemental material). His-tagged RutF was not well overexpressed, but we were able to use another highly purified flavin reductase, Fre (58), in its place for most of our studies. In the presence of both RutA and Fre (or an extract containing RutF) and the necessary flavin and pyridine nucleotide cofactors, [¹⁴C]uracil labeled at C-2 or C-6 was converted to a product with faster mobility on TLC plates (Fig. 2B). The product appeared to be more stable at pH 7 than 8.2 (data not shown). Addition of catalase to reaction mixtures to remove any H₂O₂ generated by the flavin reductase did not affect the behavior of the RutA product on TLC plates but did clear the background. [Methyl-¹⁴C]thymine was also converted to a product with faster mobility (Fig. 2C).

To identify the product produced from uracil in the RutA/F reaction, we prepared it from ¹³C- and ¹⁵N-enriched uracil. A 2D NMR ¹H-¹³C correlation spectrum in D₂O confirmed that a single product was produced (see Fig. S2A in the supplemental material). A 1D carbon spectrum showed that splitting of the ¹³C-4 signal by ¹⁵N-3 was lost in the product, whereas splitting by ¹³C-5 was retained. This indicated that the uracil ring had been cleaved between N-3 and C-4 (Fig. 3 A). We were unable to obtain ¹³C, ¹⁵N-enriched thymine commercially. However, we showed that the product from unlabeled thymine had only one new H-6-H-5 methyl correlation in a 2D ¹H total correlation spectroscopy (TOCSY) spectrum, indicating that a single product was produced and that the C-5-C-6 bond was intact (data not shown). The magnitudes of the chemical shift changes for this product were similar to those for the product from uracil, providing evidence that the two products were analogous.

• Open in new tab
• Download powerpoint

FIG. 3.

NMR evidence that RutA cleaves the uracil ring between N-3 and C-4 and incorporates O from O₂ at C-4. (A) 1D ¹³C-NMR spectrum of the C-4 resonances of the product of the RutA reaction (RutA product C4) and uracil (Uracil C4). Uracil was uniformly ¹³C/¹⁵N labeled. The C-4 resonance of uracil shows a large coupling of 65 Hz to C-5 and a small coupling of 11 Hz to N-3. The C-4 resonance of the product lacks the coupling to N-3, indicating that the bond between N-3 and C-4 was broken. The spectrum was recorded at 600 MHz. (B) The spectrum is as in panel A, except that [¹³C-4, C-5]uracil was used, the RutA reaction mixtures were bubbled with ¹⁸O₂ or ¹⁶O₂, and then equal volumes were combined. The uracil and product C-4 resonances are split by the ¹ J _{C - 4 - C - 5} coupling of 65 Hz. The product C-4 resonance also exhibits an ¹⁸O isotope shift of −0.02 ppm, indicating that oxygen was incorporated at this position. The spectrum was recorded at 800 MHz. The different chemical shifts of the species in panels A and B result from the fact that spectrum A was recorded in H₂O, whereas spectrum B was recorded in DMSO.

To further characterize the product from uracil, we prepared it from ¹³C-4, C-5-enriched uracil and a 50:50 mixture of ¹⁸O₂ and ¹⁶O₂. An isotope shift in the 1D carbon spectrum indicated that an oxygen atom derived from molecular oxygen was bound to C-4 (Fig. 3B) (16). A ¹H-¹⁵N correlation spectrum showed that there was no ¹⁸O bound to N-3 because the N-3 resonance failed to show the isotope shift of 0.138 ppm that would be expected if ¹⁸O were directly bonded to it (data not shown). Moreover, a carbon-nitrogen HSQC spectrum on product labeled with ¹³C and ¹⁵N but not ¹⁸O indicated that N-3 had been converted to NH₂ (see Fig. S2B in the supplemental material). All of the findings were in agreement with ureidoacrylate as the product.

To confirm the identity of the RutA/F product, we chemically synthesized ureidoacrylate (Z-3-ureido-2-propenoic acid) from 3-oxauracil as described in Materials and Methods. The ¹³C and ¹H shifts for the RutA/F product were the same as those for synthetic ureidoacrylate (Table 2), and the ¹H shifts and J couplings of the synthetic compound agreed well with published values (Table 3).

Finally, we obtained mass spectral data on the RutA/F product prepared from ¹⁸O₂ and a 50:50 mixture of ¹³C/¹⁵N-labeled and unlabeled uracil. Accurate mass measurements for both the ¹³C/¹⁵N and unlabeled product species were in agreement with the calculated mass values for ureidoacrylate (Fig. 4 A and 5 A). Mass spectral analysis of the RutA/F product prepared as described above also revealed a weak peak at 157.0567, which matches the theoretical mass of a ¹³C/¹⁵N-labeled species containing two atoms of ¹⁸O from molecular oxygen (within 5 ppm; theoretical value, 157.0560). Accurate mass measurements of product prepared from a mixture of ¹³C/¹⁵N-labeled and unlabeled uracil but with ¹⁶O₂ confirmed the presence of this species, which appeared to be the peracid of ureidoacrylate (Fig. 4B and 5A). In the latter case, peaks were strong enough that both the ¹³C/¹⁵N-labeled and unlabeled species containing two atoms of ¹⁶O were observed. We return to the significance of this in the Discussion.

• Open in new tab
• Download powerpoint

FIG. 4.

Mass spectrometric evidence that RutA yields ureidoacrylate and a trace of its peracid. (A) Ureidoacrylate m/z 133.0491 corresponds to [C₄H₆N₂ ¹⁶O₂ ¹⁸O + H]⁺ (calculated value, 133.0494); m/z 139.0565 corresponds to [¹³C₄H₆ ¹⁵N₂ ¹⁶O₂ ¹⁸O + H]⁺ (calculated value, 139.0569). (B) Peracid of ureidoacrylate. m/z 147.0396 corresponds to [C₄H₆N₂O₄ + H]⁺ (calculated value, 147.0400); m/z 153.0471 corresponds to [¹³C₄H₆ ¹⁵N₂O₄ + H]⁺ (calculated value, 153.0475).

• Open in new tab
• Download powerpoint

FIG. 5.

In vitro reactions catalyzed by RutA/F, RutB, and the short-chain dehydrogenase YdfG. (A) RutA/F reaction. The RutA/F reaction yields ureidoacrylate in vitro. However, there also appears to be a small amount of ureidoacrylate peracid in reaction mixtures (Fig. 4B), and we infer that this is the product of the RutA/F reaction. Mukherjee et al. (37) have evidence that the peracid is quickly reduced to ureidoacrylate by NADH spontaneously (Spont.) under conditions similar to ours (see Discussion and Fig. 6) (B) RutB reaction. The RutB reaction yields 2 mol of ammonium, HCO₃ ⁻, and malonic semialdehyde (3-oxopropionate) from ureidoacrylate. Carbamate and aminoacrylate, which hydrolyze spontaneously, are the presumed intermediates. (C) YdfG reaction. The known short-chain dehydrogenase YdfG (18) reduces malonic semialdehyde to 3-hydroxypropionic acid. In our case, malonic semialdehyde was generated from ureidoacrylate by the RutB reaction, which was run simultaneously.

The RutB reaction.RutB was initially predicted to be an isochorismatase and later a homologue of N-carbamoylsarcosine amidohydrolase (31, 32). To see whether RutB would hydrolyze ureidoacrylate, the product we obtained from the RutA/F reaction, we first used RutA and the substitute flavin reductase Fre to prepare radiolabeled ureidoacrylate from uracil. When [¹⁴C]ureidoacrylate was treated with His-tagged RutB, ¹⁴C label originating from C-2 of uracil was lost from TLC plates, as was standard [¹⁴C]HCO₃ ⁻ (data not shown). Label from C-6 smeared near the origin. The same result was obtained if the RutA, Fre, and RutB proteins were added to radiolabeled uracil simultaneously (Fig. 2B, lanes 3 and 6). Likewise, label from uracil was largely lost when cell extracts rather than purified enzymes were added to [¹⁴C-2] (data not shown) or [¹⁴C-6]uracil (Fig. 2A): at most, traces of RutA/F product were observed. When chemically synthesized ureidoacrylate was used as the substrate for RutB, approximately 2 mol of ammonium was released per mol of uracil consumed (see Fig. S3 in the supplemental material).

Hydrolytic cleavage of ureidoacrylate between N-1 and C-2 would release carbamate and aminoacrylate (Fig. 5B). The carbamate would in turn hydrolyze spontaneously to ammonium and CO₂, thus accounting for production of 1 mol of ammonium and loss of label from C-2 (46, 47). The aminoacrylate would hydrolyze spontaneously to ammonium and malonic semiadehyde, accounting for the second mole of ammonium.

To determine whether the RutB reaction released carbons 4 to 6 of the uracil ring as malonic semialdehyde, we made use of E. coli K-12 YdfG protein, a short-chain dehydrogenase that is known to oxidize 3-hydroyxypropionate and inferred to act in the reductive direction in vivo (18). (YdfG also oxidizes serine, and its best substrate is l-allo-threonine.) When RutB, YdfG, and NADPH were added to chemically synthesized ureidoacrylate simultaneously, approximately 1 mol of 3-hydroxypropionate was produced per mol of ureidoacrylate consumed, indicating that RutB did indeed release malonic semialdehyde (Table 4 and Fig. 5C). One mole of NADPH was oxidized and, as expected, approximately 2 mol of NH₄ ⁺ was released. If the RutB reaction was allowed to proceed first and then YdfG was added later, the yield of NH₄ ⁺ remained the same but the yield of 3-hydroxypropionate was greatly reduced (Table 4), possibly due to the formation of adducts by malonic semialdehyde (17). This might also account for the smearing and loss of malonic semialdehyde from TLC plates.

Requirement for the RutA, -B, and -F proteins in vivo.We constructed strains carrying nonpolar deletions in the rut genes in three backgrounds: an otherwise wild-type background; the ntrB(Con) background, in which expression of the rut operon is increased (31); and the UpBCon1 background, in which pyrimidines can be used as the sole nitrogen source at 37°C (see above). The UpBCon 1 strain (NCM4384) grows poorly at room temperature, and hence we generally studied its ability to catabolize pyrimidines at 37°C.

Strains carrying lesions in rutA or -F in any of the three backgrounds failed to grow on uridine as the sole nitrogen source (Table 5). Based on our in vitro results, this was expected for strains carrying lesions in rutA but not necessarily for strains carrying rutF lesions because RutF can be replaced in vitro by the flavin reductase Fre. Apparently no other flavin reductase can substitute for RutF in vivo. Whether this is because other flavin reductases are not present in sufficient amounts and/or do not have access to RutA or whether there is another explanation remains to be determined. In the ntrB(Con) background, where levels of Rut enzymes are elevated, addition of uridine (5 mM) to the medium inhibited growth on ammonium (5 mM) at 37°C (doubling time increased from 2 to 3 h), indicating that a toxic intermediate(s) of the Rut pathway probably accumulated. Growth inhibition persisted in an ntrB(Con) rutF strain (doubling time increased from 2 to 3 h). Uridine was not inhibitory in an ntrB(Con) rutA strain (doubling time remained 2 h in the presence of uridine), in agreement with the view that RutA is absolutely required to initiate uracil degradation. We obtained no suppressors of rutA in any background. However, we did obtain suppressors of rutF in the ntrB(Con) background. We showed that two such suppressors, which grew slowly on uridine and at different rates, released the usual 2 mol of nitrogen in utilizable form and that they excreted the usual 1 mol/mol of 3-hydroxypropionic acid into the medium (see Table S1 in the supplemental material) (31). Although we have not identified the suppressor lesions, their effects were as expected if they increased the amount or availability of another flavin reductase.

As was true of strains carrying rutA or rutF lesions, strains carrying a lesion in rutB in any of the three backgrounds we tested failed to grow on uridine as the sole nitrogen source (Table 5). As was the case for rutA strains, strains carrying a rutB lesion also failed to yield spontaneous suppressor mutations. In the ntrB(Con) background, where levels of Rut enzymes are elevated, addition of uridine to ammonium-containing medium markedly inhibited growth of the rutB strain at 37°C (data not shown), indicating that it probably accumulated a toxic intermediate(s). Based on what is known about the RutA/F and RutB reactions, this intermediate(s) would be ureidoacrylate and/or the peracid of ureidoacrylate (Fig. 5A and see the Discussion section).

Requirement for YdfG in vivo.We constructed strains carrying nonpolar deletions in ydfG in the three backgrounds described above. A strain with an insertion in ydfG in a wild-type background grew poorly on uridine as the sole nitrogen source, and an ntrB(Con) strain carrying the insertion did not grow at all (Table 5; see Fig. S4 in the supplemental material). The UpBCon1 strain carrying a ydfG insertion grew poorly on uridine at 37°C. Based on the fact that YdfG reduces malonic semialdehyde to 3-hydroxypropionic acid (see above) and therefore acts after both moles of NH₄ ⁺ have been released from the pyrimidine ring, we infer that failure of strains carrying ydfG insertions to grow well on uridine is due to toxicity of malonic semialdehyde in vivo.

Role of the RutG protein in vivo.In agreement with the bioinformatic prediction that the RutG protein was a nucleobase transporter (3, 32, 41), an otherwise wild-type strain carrying a rutG deletion failed to grow on the nucleobase uracil (0.5 mM or 2 mM) as the sole nitrogen source but used the nucleoside uridine normally. An ntrB(Con) rutG strain grew slowly on pyrimidine bases (uracil and thymine) and obtained both nitrogens from the ring. Residual growth may be accounted for by the fact that the ntrB(Con) lesion activates transcription of the gene(s) for other transporter(s) that can carry pyrimidine nucleosides/bases (59). Alternatively, or in addition, elevated expression of the rut operon in an ntrB(Con) strain may allow it to rely on a constitutively expressed transporter(s). An UpBCon1 rutG strain (NCM4384) grew well on pyrimidine bases at 37°C.

In the >20 occurrences of the rut operon outside the Enterobacteriaceae, rutG is retained only in Acinetobacter (see Table S2 in the supplemental material). Occasionally as in several methylobacteria, rutG is replaced by genes for a multisubunit transporter (ABC type) that appears to be a pyrimidine transporter and is, in other bacteria, associated with the operon for the reductive pathway of pyrimidine degradation (32, 38).

Requirement for the RutC to -E proteins in vivo.Although RutC was not required for release of ammonium in vitro, strains carrying rutC lesions in the wild-type or UpBCon1 background failed to grow on uridine as the nitrogen source (Table 5). This indicated that they probably accumulated a toxic intermediate that prevented their growth on the ammonium released from the pyrimidine ring. The ntrB(Con) rutC strain grew very slowly on uridine: it released both moles of nitrogen in utilizable form, but in contrast to its parental strain, released much less than 1 mol/mol of 3-hydroxypropionic acid into the medium (see Table S1 in the supplemental material). The latter finding indicated that RutC did not act on carbamate but probably acted on the 3-carbon intermediate released from the uracil ring (or on this portion of the molecule before hydrolysis by RutB). As explained in the Discussion, we speculate that toxicity is due to accumulation of the peracid of aminoacrylate. In the absence of RutC, cells apparently form less than the normal amount of malonic semialdehyde—and hence less 3-hydroxypropionic acid than usual—because a portion of the 3-carbon intermediate is diverted out of the Rut pathway. Although we obtained suppressors of rutC in the wild-type background, we did not identify them. Based on biochemical evidence, RutC was originally predicted to be an endoribonuclease (31, 36), but recently this has been questioned (32; see Discussion).

Like RutC, RutD was not required for release of ammonium in vitro but was required for growth on pyrimidines as the sole nitrogen source in vivo in the two backgrounds we tested (Table 5). The rutD::Kan insertion from which the original nonpolar rutD deletion was constructed may also have caused a decrease in RutC activity (see Materials and Methods) and was sufficiently toxic, even on enriched medium, that we inadvertently picked up suppressors when we introduced it into the wild-type and ntrB(Con) backgrounds. We studied one strain with a mutation that suppressed rutD in each background (NCM4088 and NCM4090, respectively). Both strains released the normal 2 mol of utilizable nitrogen from uridine but excreted much less than 1 mol/mol of 3-hydroxypropionic acid into the medium (see Table S1 in the supplemental material). This indicated that RutD, like RutC, did not act on carbamate but rather on the 3-carbon intermediate released from the uracil ring. The rutD suppressor strain NCM4088 excreted no detectable malonic acid into the growth medium (data not shown; examined as described by Loh et al. [31]), and NMR analysis of medium components failed to identify anything else excreted when it was grown on ¹³C, ¹⁵N-enriched uracil (data not shown). The rutD suppressor strain grew faster on uridine as the sole nitrogen source than its parental strain. Neither of the suppressor lesions was identified because we were not aware that they were present until we reconstructed a correct rutD deletion (rutC ⁺) in the wild-type and ntrB(Con) backgrounds late in the study (see Materials and Methods). As explained in the Discussion, we speculate that toxicity of a rutD deletion is due to the accumulation of aminoacrylate, even though it can hydrolyze spontaneously. As is the case for RutC, we think cells lacking RutD form less than the normal amount of malonic semialdehyde because a portion of the 3-carbon intermediate is diverted out of the Rut pathway.

Like RutC and RutD, RutE was required for growth on uridine in vivo, although it was not required for release of ammonium from pyrimidine rings in vitro. Wild-type or ntrB(Con) strains with a nonpolar deletion in rutE failed to grow on uridine (Table 5; see Fig. S4 in the supplemental material). Addition of uridine to ammonium-containing medium inhibited growth of the ntrB(Con) rutE strain at 37°C, confirming that this strain probably accumulated a toxic intermediate. We obtained suppressors of rutE in the ntrB(Con) background but not in the wild-type background. The two that we studied released both moles of utilizable nitrogen from uridine and excreted 1 mol/mol of 3-hydroxypropionic acid into the medium (see Table S1 in the supplemental material). The latter distinguished them from the ntrB(Con) rutC strain and from suppressors of rutD and provided evidence that they formed a normal amount of malonic semialdehyde. As explained in the Discussion, we think that the function of RutE is the same as that of YdfG: i.e., reduction of malonic semialdehyde to 3-hydroxypropionic acid (see below for identification of the lesions in rutE suppressors and the logic for this argument).

Identification of rutE suppressors and lesions that allow growth on pyrimidines at 37°C.We obtained whole-genome sequence for strains carrying rutE suppressors or lesions that allowed growth on pyrimidines at 37°C and assembled and analyzed it as described in Materials and Methods. One of the rutE suppressors (NCM4299) had a frameshift lesion early in the nemR gene that should result in truncation of the NemR protein after 65 amino acids (Table 6). (Intact NemR is 171 amino acids.) The second rutE suppressor (NCM4300), which had the same growth rate on uridine as the first, had a lesion that disrupts the inverted repeat in the binding site for NemR/RutR in the promoter-regulatory region for the nemRA operon. Finally, the UpBCon2 strain (NCM4139), which was selected spontaneously to grow on pyrimidines at 37°C but which we studied very little, also had what appeared to be a damaging lesion in nemR that converted G141 to S. The UpBCon2 strain retains good ability to grow on pyrimidines at room temperature. Hence, we were able to introduce a nonpolar rutE deletion into this strain and confirm that the NemR(G141S) lesion suppresses the loss of RutE at room temperature (Table 7; see Fig. S4 in the supplemental material). Likewise, we were able to introduce a nonpolar nemR deletion into the ΔrutE strain NCM4115 and show that it suppressed the loss of RutE at room temperature.

The NemR protein is a repressor of nemRA transcription; relief of repression apparently requires alkylation of one or more of its cysteine residues (51). The NemA gene codes for the flavoprotein N-ethylmaleimide reductase, also referred to as the “old yellow enzyme” of E. coli (55). The fact that inactivation of NemR or inactivation of its binding site at nemRA—which would increase the amount of NemR—suppressed a rutE null lesion equally well (Table 7; see Fig. S4 in the supplemental material) indicates that suppression is likely to be due to increased expression of NemA, as does the finding that NemR apparently controls only nemRA transcription, despite the fact that E. coli contains very large amounts of it (51). Presumably, high levels of N-ethylmaleimide reductase can substitute for RutE. Although both of the rutE suppressors also carried a large deletion around mioC, which encodes a mysterious FMN binding protein (Table 6) (7), this deletion was not present in the UpBCon2 strain. We found that the mioC deletion had apparently been acquired when the rutE::Kan lesion was introduced into the ntrB(Con) background (but not the wild-type background) by phage P1-mediated transduction. Based on the results presented above, the mioC deletion is not central to rutE suppression. Using markers linked to nemR by phage P1-mediated transduction, we were able to show that the NemR(G141S) lesion in the UpBCon2 strain was both necessary and sufficient for growth on pyrimidines at 37°C in the ntrB(Con) background (K.-S. Kim and W. B. Inwood, unpublished observation). However, the robust growth of UpBCon2 also required a second mutation, which we identified as an insertion of IS186 in the promoter region for the lon gene (Table 6). This insertion occurred in a hot spot and is known to eliminate Lon protease activity (42). We do not know its significance to our phenotype.

The UpBCon1 strain, which excreted a yellow compound that was identified as riboflavin (T. Mukherjee and T. Begley, personal communication), had a change in the riboswitch (called sroG) preceding the ribB (riboflavin B) gene (Table 6). The UpBCon1 strain had also acquired the deletion around mioC discussed above. Using markers linked to ribB by phage P1-mediated transduction, we showed that the sroG lesion was necessary and sufficient for growth on pyrimidines at 37°C (Kim and Inwood, unpublished). However, the robust growth of UpBCon1 also required the IS186 insertion in the lon promoter described above. It did not require the deletion around mioC.

Changes in the riboswitch for the rib operon of Bacillus subtilis resulted in riboflavin excretion by increasing transcription of the operon (35, 56). Although the effects of changing the ribB riboswitch in E. coli are less clear, we presume that the lesion we have identified increases expression of ribB at either the transcriptional or translational level. Whether directly or indirectly, this appears to result in synthesis of excess riboflavin and its excretion, although the mechanism is not obvious.

Overlap in function of RutE and YdfG.To test whether inactivation of NemR, which suppressed a rutE deletion, also suppressed a ydfG lesion, we constructed a strain carrying both nemR and ydfG lesions as described in Materials and Methods. This strain, NCM4916 [NemR(G141S) lon ntrB(Con) ΔydfG], grew faster on uridine at 37°C than a corresponding strain without the nemR mutation, NCM4714 [ntrB(Con) ΔydfG] (Table 7; see Fig. S4 in the supplemental material), providing evidence that high levels of N-ethylmaleimide reductase can substitute for the short-chain dehydrogenase YdfG (18). Likewise, strain NCM4969 [ΔnemR ΔydfG ntrB(Con)] grew faster than strain NCM4714, with which it was congenic. Thus, high levels of NemA can apparently substitute for either YdfG or RutE. (Suppression of ydfG was better at 37°C, and suppression of rutE was better at room temperature.) Suppression of both rutE and ydfG lesions by nemR lesions in turn links RutE to YdfG, whose function is known in vitro, and leads to the postulate mentioned above, namely, that RutE also reduces malonic semialdehyde to 3-hydroxypropionic acid.

The requirement for YdfG function or RutE function for utilization of uridine at 37°C was decreased in the UpBCon1 background (i.e., in the presence of the sroG lesion) (Table 7; see Fig. S4 in the supplemental material). This hints that large amounts of reduced flavin may also be able to drive reduction of malonic semialdehyde in vivo, either per se or through an unidentified enzyme(s).

Evidence that RutE and YdfG have the same function.The RutE protein is predicted to belong to nitroreductase-like subfamily 5, which contains proteins of unknown function (9, 27, 32). Like members of the greater nitroreductase family, RutE is believed to use FMN as a cofactor. It is required for growth on uridine as the sole nitrogen source in both the wild-type and ntrB(Con) backgrounds (Table 7; see Fig. S4 in the supplemental material). Genetic evidence indicates that RutE has the same function as YdfG: i.e., both reduce malonic semialdehyde to 3-hydroxypropionic acid, although presumably by different mechanisms. Furthermore, the evidence indicates that toxicity of malonic semialdehyde, not the rate of release of ammonium, limits growth of E. coli K-12 on pyrimidines as the sole nitrogen source at high temperatures. The reasoning for these conclusions is as follows. First, relief of transcriptional repression of nemA, which codes for N-ethylmaleimide reductase, the “old yellow” enzyme of E. coli, suppresses the absence of either RutE or YdfG in vivo (Table 7; see Fig. S4 in the supplemental material). This would be expected if all three had the same biochemical function. In agreement with this view, overproduction of N-ethylmaleimide reductase in a rutE null strain results in excretion of the usual 1 mol/mol of 3-hydropropionic acid into the growth medium (see Table S1 in the supplemental material). Finally, overexpression of N-ethylmaleimide reductase allows an ntrB(Con) strain, which expresses the rut operon at high levels, to grow on pyrimidines as the sole nitrogen source at 37°C (although not as well as the UpBCon2 strain; see Results). Additional lines of bioinformatic evidence support the view that RutE catalyzes reduction of malonic semialdehyde to 3-hydropropionate. First, the rutE gene is often absent from the rut operon (see Table S2 in the supplemental material), in agreement with the view that RutE acts after the nitrogens have been extracted from pyrimidine rings. Second, the five rut operons in Acinetobacter genomes (18 genomes total), all of which lack rutE, have a gene that codes for an enzyme in the same superfamily as YdfG and is predicted to reduce malonic semialdehyde and 2-methyl malonic semialdehyde to their corresponding alcohols (32). Finally, the rut operons in the two Alteromonas species for which whole genome sequences are available contain not only rutE but also the gene for an additional enzyme predicted to detoxify malonic semialdehyde by oxidizing it rather than reducing it (malonate semialdehyde/methyl malonate semialdehyde dehydrogenase, which would oxidize malonic semialdehyde to acetyl-S-coenzyme A [CoA]) (32, 50) (see Table S2 in the supplemental material). Neither genome carries a ydfG gene. Biochemical studies of RutE will be particularly interesting because flavoenzymes generally participate in oxidation of alcohols rather than reduction of aldehydes (24).

Speculations on RutD and RutC function.Like other Rut pathway proteins, both RutD and RutC are required for growth on uridine as the sole nitrogen source, despite the fact that they are not required for release of ammonium in vitro. We speculate that RutD, a hypothetical α/β-hydrolase with no close relatives (32), increases the rate of spontaneous hydrolysis of aminoacrylate to malonic semialdehyde (Fig. 6). This would be analogous to the role of carbonic anhydrases in accelerating the rate of spontaneous hydration of CO₂.

Finally, we speculate that RutC, a member of a family of proteins without a clearly defined function (32), reduces the peracid of aminoacrylate to aminoacrylate, the substrate for RutD (Fig. 6). Members of the RutC family appear to bind toxic metabolic intermediates (10, 14). Both of the other members of the family in E. coli, TdcF (threonine deaminase catabolic F) and YjgF, appear to be involved in metabolism of the toxic intermediate 2-ketobutyrate (8, 10, 14, 29). Structurally, RutC family proteins are trimers with binding clefts for small ligands at monomer interfaces (53). In the clefts, they carry an invariant R that is often followed by XC. The structure of the E. coli TdcF protein has been determined with 2-ketobutyrate bound: oddly, it was bound as the rare enol tautomer, with its carboxylate group doubly hydrogen bonded to the guanidinium group of the invariant R and its enol OH group bonded to both the backbone amide of the conserved C and the carboxyl group of the conserved E120 in the adjacent subunit (8) (see Fig. S5 in the supplemental material). The side chain of the corresponding C in the E. coli YjgF protein—whose structure has been determined only in its unliganded form—was derivatized with what appeared to be a thiophosphate or a thiosulfate (53). By analogy to what is known about TdcF and YjgF, we postulate that RutC binds the peracid form of aminoacrylate (see Fig. S5 in the supplemental material) and that it may use the XC¹⁰⁷XXC¹¹⁰ motif adjacent to its invariant R¹⁰⁵ to reduce the peracid to the carboxylic acid (aminoacrylate). If RutC binds aminoacrylate peracid in its stable amine form—as would be predicted—it would also inhibit spontaneous hydrolysis of the amino group. When aminoacrylate was released by RutC, RutD could then increase its spontaneous rate of hydrolysis by catalyzing formation of the rare imine tautomer. The roles of RutC and RutD would be to insure that reduction of the peracid form of aminoacrylate and hydrolysis of aminoacrylate occur rapidly and in a particular order. These admittedly speculative ideas provide a framework for further biochemical and genetic studies. In the latter connection, it will be interesting to determine the identities of rutC and rutD suppressors, for which tools are now available (see Materials and Methods), and to understand why a rutC strain and rutD suppressors go off the pathway and excrete less than the usual amount of 3-hydroxypropionic acid into the medium (see Table S1 in the supplemental material). Apparently, they generate less than the usual amount of toxic malonic semialdehyde (see Results).

Conclusions.In summary, Rut pathway enzymes oxidatively cleave the pyrimidine ring to produce a series of reactive, toxic intermediates that includes strong oxidizing agents (peracids) and compounds known to polymerize readily (ureidoacrylates and aminoacrylates) or form adducts (malonic semialdehyde) (Fig. 6). Only half of the Rut enzymes (RutA, RutF, and RutB) are required in vitro to release both nitrogens from the pyrimidine ring as NH₄ ⁺ (Fig. 5). The other half (RutC, RutD, and RutE) are nonetheless required in vivo, apparently to prevent accumulation of toxic intermediates and by-products. We postulate that they act in order on the 3-carbon intermediate released by RutB (Fig. 6). The function of RutE overlaps with that of the short-chain dehydrogenase YdfG (Tables 6 and 7; see Fig. S4 in the supplemental material), and both are required in vivo: apparently neither alone has sufficient activity, and the two together are still not sufficient for growth at 37°C.

Although E. coli does not grow on pyrimidines as the sole nitrogen source at 37°C, it transcribes the rut operon very highly at this temperature under nitrogen-limiting conditions (31, 59). Presumably the Rut pathway allows E. coli to use pyrimidines, which are readily available degradation products of RNA, as part of the nitrogen source at 37°C. (The YdfG protein, which is coded for outside the rut operon, may not be required under these circumstances.) Forcing their use as the sole nitrogen source at any temperature is a trick of the experimentalist. The rut operon is highly expressed, even in the absence of exogenous pyrimidines. Whether the Rut pathway is also used to decrease the internal free pool concentrations of pyrimidines under nitrogen-limiting conditions and/or to generate toxic intermediates that help slow the growth of E. coli in a coordinated way are intriguing possibilities that remain to be explored.