The genome of A. flavithermus strain WK1 consists of a single, circular chromosome of 2,846,746 bp (Figure 1) with an average G+C content of 41.78% (Table 1). The genome encompasses 2,863 predicted protein-coding genes, 8 rRNA (16S-23S-5S) operons, 77 tRNA genes, and 19 predicted riboswitches. Of the 2,863 predicted proteins, 1,929 have been assigned probable biological functions, 418 were conserved proteins with only general function predicted, and for 516 putative proteins no function was predicted (of these, 110 proteins had no detectable homologs in the NCBI protein database). The genome contains one prophage region with 44 genes (Aflv_0639-0682) and encodes 105 transposases. In its gene order and the phylogenetic affinities of the encoded proteins, A. flavithermus WK1 is a typical member of the family Bacillaceae, with Geobacillus kaustophilus and Geobacillus thermodenitrificans as its closest neighbors (see below). Pair-wise genome alignments show high conservation of gene order between A. flavithermus, G. kaustophilus and B. subtilis (Figure 2). Anoxybacillus flavithermus WK1 has a typical firmicute proteome, with 89% of the predicted open reading frames (ORFs) having closest homologs in Bacillus spp. (Figure S1 in Additional data file 1). However, the A. flavithermus WK1 genome is the smallest among the sequenced members of Bacillaceae and generally encodes fewer paralogous proteins than other bacilli (Table S1 in Additional data file 1).

Despite its much smaller genome size, A. flavithermus appears to retain most of the key metabolic pathways present in B. subtilis and other bacilli. It has a complete set of enzymes for biosynthesis of all amino acids, nucleotides and cofactors, with the sole exception of the molybdenum cofactor (Table S2 in Additional data file 1). Cells of A. flavithermus had been originally reported to reduce nitrate 46; however, in subsequent work, nitrate reductase activity has not been observed in this organism 15. In accord with the latter report, the A. flavithermus WK1 genome encodes neither the assimilatory nitrate/nitrite reductase complex (NasBCDE) nor the respiratory nitrate reductase complex (NarGHJI), both of which are present and functional in B. subtilis 1617, nor the third (proteobacterial) type of nitrate reductase (NapAB) 18. Nitrate/nitrite transporters NasA and NarK are missing in A. flavithermus as well. The loss of nitrate reductases in A. flavithermus WK1 appears to be a recent event, given that G. kaustophilus encodes the assimilatory nitrate reductase, whereas G. thermodenitrificans encodes the respiratory nitrate reductase complex. In accordance with the loss of nitrate reductases, A. flavithermus WK1 has lost the entire set of enzymes involved in the biosynthesis of the molybdenum cofactor of nitrate reductase, as well as the molybdate-specific ABC (ATP-binding cassette)-type transporter, all of which are encoded in G. kaustophilus and G. thermodenitrificans. Molybdenum-dependent xanthine dehydrogenase and its homologs YoaE (putative formate dehydrogenase) and YyaE have been lost as well. As suggested in 19, the loss of molybdate metabolism could be part of a strategy to avoid generation of reactive oxygen species.

As the name suggests, members of the genus Anoxybacillus were initially described as obligate or facultative anaerobes 45. However, the initial description of (Anoxy)bacillus flavithermus already mentioned its capability to grow in aerobic conditions 6. Examination of the A. flavithermus WK1 genome revealed that it encodes an electron transfer chain that is as complex as that of B. subtilis and appears to be well-suited for using oxygen as terminal electron acceptor. The electron transfer chain of A. flavithermus includes NADH dehydrogenase, succinate dehydrogenase, quinol oxidases of bd type and aa₃ type, menaquinol:cytochrome c oxidoreductase and cytochrome c oxidase, as well as two operons encoding the electron transfer flavoprotein (Table 2). Anoxybacillus flavithermus also encodes a variety of enzymes that are important for the defense against oxygen reactive species, such as catalase (peroxidase I), Mn-containing catalase, Mn-, Fe-, and Cu,Zn-dependent superoxide dismutases (the latter, in contrast to B. subtilis YojM, has both Cu-binding histidine residues), thiol peroxidase, and glutathione peroxidase (Table 2). The presence of these genes in the genome suggests that A. flavithermus WK1 should be able to thrive in aerobic conditions. Indeed, isolation of this strain, similarly to the type strain A. flavithermus DSM 2641, has been carried out in open air, without the use of anaerobic techniques 6920.

Anoxybacillus flavithermus WK1 grows well anaerobically in rich media, such as tryptic soy broth (TSB). Owing to the absence of nitrate and nitrite reductases (see above), its anaerobic growth cannot rely on nitrate or nitrite respiration and apparently proceeds by fermentation. Fermentative growth of B. subtilis requires phosphotransacetylase, acetate kinase and L-lactate dehydrogenase genes 13. All these genes are conserved in A. flavithermus (pta, Aflv_2760; ack, Aflv_0480; lctE, Aflv_0889), suggesting that, like B. subtilis, this bacterium can ferment glucose and pyruvate into acetate 1. However, catabolic acetolactate synthase AlsSD and acetolactate dehydrogenase, which are responsible for acetoin production by fermenting B. subtilis 1, are missing in A. flavithermus, indicating that it cannot produce acetoin.

In agreement with the experimental data 6, genome analysis indicates that A. flavithermus is able to utilize a variety of carbohydrates as sole carbon sources. It has at least four sugar phosphotransferase systems with predicted specificity for glucose, fructose, sucrose, and mannitol. Additionally, it encodes ABC-type transporters for ribose, glycerol-3-phosphate, and maltose, and several ABC-type sugar transporters of unknown specificity. A complete set of enzymes was identified for general carbohydrate metabolism (glycolysis, the TCA cycle, and the pentose phosphate pathway, but not the Entner-Doudoroff pathway). The A. flavithermus genome also contains a gene cluster (Aflv_2610-2618) that is very similar to the gene cluster associated with antibiotic production and secretion in many other Gram-positive bacteria 21, suggesting that A. flavithermus might be able to produce bactericidal peptides. It is not obvious which of these systems are relevant to the survival of A. flavithermus in silica solutions, but they might facilitate its growth in powdered milk and similar habitats.

In a phylogenetic tree constructed using a concatenated alignment of the RNA polymerase subunits RpoA, RpoB, and RpoC, A. flavithermus, G. kaustophilus, and G. thermodenitrificans grouped together and formed a deep branch within the Bacillus cluster (Figure 3). A distinct Anoxybacillus/Geobacillus branch is also seen in a gene content tree that was constructed on the basis of the presence or absence of particular protein families in the genomes of 26 species of firmicutes and 2 actinobacteria (used as an outgroup; Figure S2 in Additional data file 1).

Anoxybacillus flavithermus WK1 has a relatively small genome compared to other Bacillus species. To determine which genes were likely to have been lost and gained in this lineage, we reconstructed the most parsimonious scenario of evolution 22 from the last common ancestor of the firmicutes. The reconstruction was performed on the basis of the assignment of A. flavithermus to the Clusters of Orthologous Groups of proteins (COGs), followed by the comparison of COG-based phyletic patterns of 20 other bacilli, 5 clostridia, and 6 mollicutes. This approach assigned 2,015 genes (COGs) to the common ancestor of A. flavithermus and G. kaustophilus (Figure 4). The reconstruction results suggest that a massive gene loss (-437 genes) occurred during evolution from the common ancestor of Bacillaceae to the common ancestor of Anoxybacillus and Geobacillus. The majority of the genes shared between A. flavithermus and G. kaustophilus are also shared with other Bacillus species. Gene losses in the Geobacillus/Anoxybacillus branch include, among others, genes encoding the nitrogen regulatory protein PII, ABC-type proline/glycine betaine transport system, methionine synthase II (cobalamin-independent), sorbitol-specific phosphotransferase system, β-xylosidase, and some dTDP-sugar metabolism genes (Table S3 in Additional data file 1). However, 62 gene gains were inferred as well, including several genes coding for cobalamin biosynthesis enzymes, methylmalonyl-CoA mutase, genes involved in assembly of type IV pili (Aflv_0630-0632), an uncharacterized ABC-type transport system, and 16 genes encoding uncharacterized conserved proteins (Table S3 in Additional data file 1). After the split of the Anoxybacillus and Geobacillus lineages, A. flavithermus continued to show strong genome reduction (-292 genes) compared to G. kaustophilus (-124 genes), losing, in particular, some genes of nitrogen and carbohydrate metabolism. In addition, A. flavithermus has apparently experienced less gene gain (+88) than G. kaustophilus (+158). The few genes likely acquired in the Anoxybacillus lineage include the clustered regularly interspaced short palindromic repeat (CRISPR)-associated genes (Aflv_0764-0771) that form an antisense RNA-based system of phage resistance, which is often associated with thermophily 2324.

Being a free-living environmental microorganism, A. flavithermus encodes numerous proteins involved in signal transduction. These include 23 sensor histidine kinases and 24 response regulators (16 pairs of which are clustered in operons), 20 methyl-accepting chemotaxis proteins, 5 predicted eukaryotic-type Ser/Thr protein kinases, and 21 proteins involved in metabolism of cyclic diguanylate (cyclic (3',5')-dimeric guanosine monophosphate (c-di-GMP)), a recently recognized secondary messenger that regulates transition from motility to sessility and biofilm formation in a variety of bacteria 25. Compared to other bacilli, this set is significantly enriched in chemotaxis transducers and c-di-GMP-related proteins 26. Anoxybacillus flavithermus encodes 12 proteins with the diguanylate cyclase (GGDEF) domain, 6 of which also contain the c-di-GMP phosphodiesterase (EAL) domain, and one combines GGDEF with an alternative c-di-GMP phosphodiesterase (HD-GYP) domain. Anoxybacillus flavithermus WK1 also encodes two proteins with the EAL domain and seven proteins with the HD-GYP domain that do not contain the GGDEF domain. In addition, it encodes two proteins with the PilZ domain 27, which serves as a c-di-GMP-binding adaptor protein 2829. The total number of proteins implicated in c-di-GMP turnover in A. flavithermus is third highest among all Gram-positive bacteria sequenced to date, after Clostridium difficile and Desulfitobacterium hafniense, which have much larger genomes 2630.

The abundance of c-di-GMP-related proteins suggests that regulation of biofilm formation plays an important role in the physiology of A. flavithermus. Indeed, scanning electron microphotographs of A. flavithermus cells cultured in the presence of high amounts of silica showed that the presence of biofilm had a major effect on the form of silica precipitation. In the absence of bacteria, the prevailing mode of silica precipitation was the formation of a layer of amorphous silica nanospherules (Figure 5a). In the presence of bacteria, silica precipitates were often associated with individual cells of A. flavithermus (Figure 5b), suggesting that these cells might serve as nucleation sites for sinter formation. However, in the culture of A. flavithermus cells attached as a biofilm to a glass slide, silica precipitates were mostly bound to the exopolysaccharide material of the biofilm (Figure 5c,d). Biofilm-associated silica was often seen forming extensive granular silica precipitates (Figure 5e). Further incubation led to the development of a complex, multi-layered biofilm that was impregnated with silica particles (Figure 5f). Obviously, A. flavithermus biofilm formation played a key role in determining the structural nature of the silica sinter. Indeed, A. flavithermus WK1 retains some of the genes (Table 3) that are required for biofilm formation in B. subtilis 3132. Proteins encoded by these genes include: the master regulators of biofilm formation AbrB (Aflv_0031) and SinR (Aflv_2245); α-phosphoglucomutase YhxB (Aflv_2333), which is probably involved in exopolysaccharide synthesis; EcsB (Aflv_2284), the membrane subunit of an ABC-type transporter that could promote secretion of protein components of the extracellular matrix; an HD-superfamily hydrolase YqeK (Aflv_0816) that is required for the formation of thick pellicles; YlbF (Aflv_1855), a positive regulator of competence factor ComK; and YmcA (Afla1522), a protein of unknown function. Other biofilm-forming proteins of B subtilis, namely, the AbrB- and SinR-regulated genes tasA (yqhF) or yqfM 3334, are absent in the smaller genome of A. flavithermus.

The existence of c-di-GMP-mediated signal transduction pathways also suggested that biofilm formation in A. flavithermus could be regulated in response to environmental conditions. To investigate possible mechanisms of silica adaptation, we compared protein expression profiles of A. flavithermus in the presence and absence of silica using two-dimensional electrophoresis and matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) mass spectrometry analyses (Figure S3 in Additional data file 1). Although samples from three independent experiments showed significant variance and the expression changes could not be statistically proven (Table S4 in Additional data file 1), the trends that they revealed provided certain clues to the A. flavithermus adaptation to silica. After exposure of batch cultures to 10.7 mM (300 ppm) silica (a mixture of monomeric H₄SiO₄ and polymerized silicic acid 35) for 8 hours, expression of 19 proteins was increased at least 1.5-fold in each of three independent experiments, whereas expression of 18 proteins was found to be decreased (Table S4 in Additional data file 1). Most of these proteins were products of house-keeping genes whose up- or down-regulation could be related to the general stress in the presence of silica, as suggested by the increased expression of the alkaline shock protein Asp23 (Aflv_1780) and the carboxylesterase YvaK (Aflv_2499), which are stress-induced in B. subtilis 36. The increased expression of AbrB (Aflv_0031), a key transcriptional regulator of biofilm-related genes in B. subtilis, suggested that exposure to silica could, indeed, trigger biofilm formation by A. flavithermus. Of particular interest was the differential effect of silica on the expression of two close paralogs, putrescine aminopropyltransferase (spermidine synthase) SpeE (Aflv_2750) and SpeE-like protein Aflv_1437. Expression of SpeE, which is part of the polyamine biosynthesis pathway of B. subtilis 37, was suppressed by exposure to silica. In contrast, SpeE-like protein Aflv_1437, which could participate in the synthesis of some other polyamine(s) (see, for example, 38), was up-regulated (Table S4 in Additional data file 1). A predicted arginase (Aflv_0146), which catalyzes the first step in the synthesis of putrescine (the substrate of SpeE), namely, conversion of arginine to ornithine (Figure 6), was also up-regulated, whereas the expression of predicted arginine decarboxylase (Aflv_1886) and agmatinase (Aflv_2749), which comprise an alternative route for the synthesis of putrescine, was very low and, apparently, remained unchanged (data not shown), suggesting that putrescine was primarily produced via the arginase route. Given that long-chain polyamines (LCPAs) are crucial in the formation of silica nanostructures in diatoms 3940414243, these data suggested a link between polyamine biosynthesis and biofilm formation in A. flavithermus. As a first step towards characterizing this link, proteins encoded by genes Aflv_0024, Aflv_0146, Aflv_1437, Aflv_1886, Aflv_2749, and Aflv_2750 were individually expressed, purified, and confirmed to function as, respectively, ornithine decarboxylase, arginase, spermine synthase, arginine decarboxylase, agmatinase, and spermidine synthase (Figures S4-S6 in Additional data file 1). In the general route, spermine synthase converts spermidine into spermine by transferring an aminopropyl group. The spermine synthase (Aflv_1437) identified here converts putrescine directly into spermine by adding two aminopropyl groups, raising the possibility of the formation of longer chain polyamines by sequentially adding multiple aminopropyl groups. The proposed roles of these enzymes in LCPA biosynthesis in A. flavithermus are shown in Figure 2.

We also examined protein expression profiles in the cells grown in the presence or absence of silica for 7 days. Sinters started forming in the silica-containing sample 5 days after inoculation, so by the end of the incubation the cells became silicified. Owing to the problems with collecting and analyzing silicified A. flavithermus cells, no attempt has been made to replicate this experiment, so these results were only considered in comparison to the samples from 8-hour exposure to silica. Spermine synthase Aflv_1437 was not detected in either silicified or control cells (last column of Table S4 in Additional data file 1), and arginase (Aflv_0146; Figure S7 in Additional data file 1) was only detected in the silicified cells at very low abundance. In contrast, spermidine synthase Aflv_2750 was detected at similar levels in both types of cells, indicating general cellular functions for spermidine. Remarkably, the transcriptional regulator AbrB (Aflv_0031) remained moderately up-regulated in the silicified cells, suggesting that it might play a general role in silica adaptation of A. flavithermus. Also up-regulated in both silica conditions were chemotaxis response regulator CheY (Aflv_1727), thiol peroxidase Aflv_0478, which is apparently involved in antioxidant defense, and methionine aminopeptidase Aflv_0127. Those proteins could also play a role in silica niche adaptation of A. flavithermus.

The genome of A. flavithermus was sequenced using the whole-genome-shotgun approach as previously described 60, using genomic DNA that was randomly sheared to generate 3 kb and 6 kb fragments. These fragments were size-selected on agarose gels, purified, end-repaired, ligated to pUC118 vectors, and transformed into DH10B competent cells by electroporation. Plasmids from positive clones were sequenced using Beckman CEQ 8000 (Beckman Coulter, Fullerton, CA, USA) and ABI 3730xl (Applied Biosystems, Foster City, CA, USA) sequencers. A total of 55,975 valid sequences were used for assembly with PHRED/PHRAP/CONSED 61, CAP3 62, and SEQMAN II (DNAStar) programs. Further 3,863 sequences were used to close gaps between contigs and to improve overall sequence quality of contigs. Long PCR reactions were performed to verify sequence assembly. Protein-coding genes were predicted using GLIMMER 63 followed by BLASTX 64 searches of intergenic regions between predicted ORFs. Transfer RNAs were predicted by tRNAscan-SE 65. Genome annotation was performed by running BLAST and PSI-BLAST against the NCBI protein database and the COG database with manual verification as described previously 60. Metabolic pathways were analyzed by comparing COG assignments of A. flavithermus proteins with the standard sets of COGs involved in each pathway 66. Phylogenetic analysis was performed as described 67.

Biofilm formation by A. flavithermus cells grown in the presence of silica was studied by incubating the cells in a chemostat-like system, consisting of a 500 ml serum vial, capped with a rubber seal with two input and one output lines. This was filled halfway with 300 ml of TSB. The two input lines were fed through the rubber seal and connected via peristaltic pumps to sterile reservoirs. One reservoir contained 2 × TSB and the other water. They were each fed at 0.15 ml per minute giving 1 × TSB in the vial. An output line connected to another peristaltic pump maintained the medium level at 300 ml. A final output line was fit with a luer valve and syringe to allow samples to be removed from the reservoir. A glass slide stood upright in the vial as a substrate for silica sinters, to be observed by scanning electron microscopy at the conclusion of the experiment. The cultivation vessel was contained in a 60°C oven and shaken gently at 100 rpm to simulate wave motion. After running the system for two days to ensure it was sterile, the medium was inoculated with A. flavithermus WK1 through the luer-fitted line. The system ran for two days to build up cell mass, then samples of 200 ml were taken for three successive days. Samples were centrifuged, the pellet washed 3 times in buffer (68 mM NaCl, 3 mM KCl, 1.5 mM KH₂PO₄, 9 mM NaH₂PO₄, 50 mM TRIS, pH 8.0), stored at -20°C and later freeze-dried. The system was running at pH 5.8 and OD₆₀₀ 0.15 during this time. After three days, the water was replaced with 1,000 mg/kg silica solution adjusted to pH 7. This flowed through a 200°C oven before reaching the cultivation vessel in order to monomerize the silica and sterilize the water. Samples were taken after 1, 2, 3 and 7 days and prepared as above. As time progressed, there was increasingly more solid, amorphous silica in the vessel, as this was not removed by the outflow. The system remained at pH 5.8 but there was no longer any way to reliably measure OD₆₀₀ because of the silica precipitate. At the end of incubation, some of the amorphous silica and the slide were removed to be fixed in 2% glutaraldehyde. Samples for scanning electron microscopy were removed from storage and allowed to air-dry before coating with gold/palladium. Scanning electron microscopy examination was done on a Hitachi S-800 Field Emission scanning electron microscope operating at 15 kV.

Anoxybacillus flavithermus cells were grown in TSB at 60°C with shaking at 200 rpm on an orbital shaker (Thermo Electron Co., Waltham, MA, USA) to OD₆₀₀ of 0.6, followed by the addition of silica to a final concentration of 10.7 mM and growth for another 8 hours. The same batch of the culture without added silica served as the control. Cells were harvested by centrifugation at 10,000 × g at 4°C for 10 minutes, extracellular proteins from the supernatant were collected and cellular proteins from the pellet were solubilized 68. Immunoelectrophoresis (the first dimension) was carried out on IPG strips (Amersham Pharmacia Biotech, Uppsala, Sweden) in a Multiphor II electrophoresis unit (Amersham Pharmacia Biotech) with running conditions as described by Büttner et al. 69. For the second dimension, vertical slab SDS-PAGE (12%) was run in a Bio-Rad Protean II Xi unit (Bio-Rad Laboratories, Hercules, CA, USA). Gels were stained with colloidal CBB G-250 70, and scanned with a PowerLook 1000 (UMAX Technologies Inc., Fremont, CA, USA). PDQuest version 7.3.0 (Bio-Rad) was used for image analysis. Proteins were classified as being differentially expressed under the two conditions when spot intensity showed at least 1.5-fold change.

For protein identification, spots were excised from the gels, washed with 25 mM NH₄HCO₃ in 50% (v/v) acetonitrile for 3 × 15 minutes at room temperature, dried in a vacuum centrifuge, and incubated in 50 μl digestion solution consisted of 25 mM NH₄HCO₃ in 0.1% acetic acid and 12.5 ng/mL of trypsin (Promega, Madison, WI, USA) at 37°C overnight. The digested protein (0.3 μl) was spotted on a MALDI sample plate with the same volume of matrix (10 mg/ml α-cyano-4-hydroxycinnamic acid in 50% acetonitrile, 0.1% trifluoroacetic acid). Peptide mass spectra were obtained on a MALDI-TOF/TOF mass spectrometer (4700 Proteomics Analyzer, Applied Biosystems) in the positive ion reflector mode. The mass spectrometry spectra were internally calibrated with a mass standard kit for the 4700 Proteomics Analyzer. Proteins were identified by automated peptide mass fingerprinting using the Global Proteome Server Explorer™ software (Version 3.5, Applied Biosystems) against an in-house sequence database of A. flavithermus proteins. Peak lists (S/N > 10) were extracted from raw data for the data processing, and positive identifications were accepted up to 95% of confidence level. The following criteria were used for the database searches: maximum one missed cleavage per peptide; mass tolerance of 0.1 Da, and the acceptation of carbamidomethylation for cysteine and oxidation for methionine.

The genes Aflv_0024, Aflv_1886, Aflv_2749, and Aflv_1437 were cloned into the pET-14b vector, Aflv_0146 into pET-3a, and Aflv_2750 into pET-28a. Escherichia coli strain BL21 carrying each of the recombinant plasmids was grown overnight with shaking at 37°C in Luria broth containing 100 mg/ml ampicillin. The overnight culture (4 ml) was inoculated into 400 ml of fresh Luria broth and grown to mid-log phase (A₆₀₀ = 0.6). Expression of Aflv_0146, Aflv_1437, Aflv_1886, Aflv_2749, and Aflv_2750 products was induced with 0.1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at 37°C for 4 hours, and expression of the Aflv_0024 product was induced with 0.1 mM IPTG at 12°C for 8 hours. After IPTG induction, the cells were harvested by centrifugation at 6,000 × g at 4°C for 5 minutes, washed with binding buffer (10 mM imidazole, 300 mM NaCl and 50 mM Tris-HCl, pH 8.0), resuspended in 5 ml of binding buffer containing 1 mM phenylmethylsulfonyl fluoride and 1 mg/ml of lysozyme, and sonicated for 10 1-minute cycles with 1-second pulse on alternating 1-second pulse off at 95% of the maximum power (200 W) using an UP200S Ultraschallprozessor with a tapered microtip. The lysate of Aflv_1886 was further incubated for 10 minutes at 60°C. After centrifugation at 12,000 × g at 4°C for 30 minutes, the crude extract containing 6× His-tagged fusion proteins was purified by nickel ion affinity chromatography with a Chelating Sepharose Fast Flow column (GE Healthcare, Piscataway, NJ, USA) according to the manufacturer's instructions. The column was washed successively with 100 ml of wash buffer (25 mM imidazole, 300 mM NaCl, and 50 mM Tris-HCl, pH 8.0), and the fusion proteins were eluted with the elution buffer (250 mM imidazole, 300 mM NaCl and 50 mM Tris-HCl, pH 8.0), and dialyzed in 0.1 M Tris-HCl buffer (pH 8.8). Protein concentration was determined by the Bradford method. For SDS-PAGE, proteins were denatured at 100°C for 5 minutes in the presence of 0.1% SDS and 1% 2-mercaptoethanol, loaded in a 5% (w/v) stacking gel and separated in a 10% (w/v) separation gel. The gel was stained with Coomassie Bright Blue R250. The molecular weight markers were from the LMW-SDS Marker Kit (GE Healthcare).

Reactions catalyzed by arginase, arginine decarboxylase, ornithine decarboxylase, agmatinase, spermidine synthase and spermine synthase were carried out as previously described 71727374. The activities of arginase and arginine decarboxylase were determined by thin-layer chromatography 75. The activities of the other enzymes were assayed by high-performance liquid chromatography (HPLC) after Schotten-Baumann benzoylation as previously described 747677. HPLC analysis was performed with a Venusil XBP C18 column (4.6 × 250 mm) in conjunction with a LC-20AT (Shimadzu, Kyoto, Japan) HPLC apparatus. Benzoyl putrescine, spermidine and spermine were eluted by a gradient started with 60% methanol in water and proceeded linearly to 100% methanol, with a flow rate of 0.8 ml/minute over 20 minutes, and detected at a wavelength of 229 nm.