Combined metabolomic and genetic approaches reveal a link between the polyamine pathway and albumin 2 in developing pea seeds 1

Several legume seed proteins that are potentially allergenic, poorly digested by farm animals, and/or have undesirable functional properties, have been described. One of these is the albumin protein in pea ( Pisum sativum L.) called PA2. A naturally occurring mutant line that lacks PA2 has been exploited in studies to determine the biological function of this non-storage protein in seed development. The mutant, which has a small seed, a tall plant phenotype, and lacks most of the PA2-encoding genes, has been crossed with a standard cultivar ( cv. ) Birte that contains PA2 to give rise to a recombinant inbred (RI) population. An F 3 line carrying the mutation, and having a short plant phenotype, has been used to generate back-cross (BC) lines with the cv. Birte. Despite having a lower albumin content, seeds from the mutant parent and RI lines lacking PA2 have an equivalent or higher seed nitrogen content. Metabolite profiling of seeds revealed major differences in amino acid composition and polyamine content in the two parent lines. This was investigated further in BC lines, where the effects of differences in seed size and plant height between the two parents were eliminated. Here, differences in polyamine synthesis were maintained as was a difference in total seed protein between the BC line lacking PA2 and cv. Birte. Analysis of enzyme activities in the pathways of polyamine synthesis revealed that the differences in spermidine content were attributable to changes in the overall activities of spermidine synthase and arginine decarboxylase. Although the genes encoding spermidine synthase and PA2 both localised to the pea linkage group I, the two loci were shown not to be closely linked and to have recombined in the BC lines. A distinct locus on linkage group III contains a gene that is related to PA2 but expressed predominantly in flowers. The results provide evidence for a role of PA2 in regulating polyamine metabolism, which has important functions in development, metabolism and stress responses in plants.


INTRODUCTION
Pea albumin 2 (PA2) is a seed protein that has been shown to be resistant to digestion in piglets and chickens and can persist until the end of the digestive tract (Crévieu et al., 1997;Salgado et al., 2003;Le Gall et al., 2007). Studies of the equivalent protein in chickpea (Cicer arietinum L.) have shown that both chickpea and pea proteins behave as potential allergens in humans (Vioque et al., 1998). Sera from chickpea-sensitive individuals have been shown to give positive reactions against PA2 from both legumes, and both proteins were shown to agglutinate papainized erythrocytes (Vioque et al., 1998).
Studies of the physicochemical properties of pea PA2 suggested that a free sulphydryl group could lead to polymerisation, through disulphide interchange, and partial insolubility of protein isolates (Gruen et al., 1987). Taken together, these data suggest that a reduced content of PA2 in pea seeds would lead to a number of improved seed quality characteristics.
The biological relevance of PA2 is unclear. It is not a classic storage protein, lacks a signal peptide, and is not degraded on germination (Higgins et al., 1987). The protein is composed of four copies of a conserved repeat, described as a hemopexin-type repeat, and is thus related structurally to a group of mammalian regulatory proteins, that includes vitronectin (Jenne, 1991). It has been shown that chickpea PA2 binds hemin, in agreement with this relationship (Pedroche et al., 2005). It is possible that PA2 binds other small ligands, including thiamine (Adamek-Świerczyńska and Kozik, 2002).
These data suggest that PA2 may act as a regulatory protein in controlling biological processes dependent on ligand availability.
In this work, a naturally occurring pea mutant that lacks this protein has been identified through germplasm screening. Despite having a lower albumin content, mutant seeds appear to have an elevated nitrogen content under a range of growth conditions. So, provided there are no negative pleiotropic effects, this mutation may have value for crop improvement. The mutation, which is in a wild uncultivated line, was shown to be due to a deletion of most 6 of the structural genes encoding PA2. Genetic crosses involving the mutant line were established, and metabolomic approaches employed to identify biochemical changes during seed development that may be linked directly or indirectly to the mutation.
The results indicate that metabolic changes occur in the polyamine pathway with altered amounts of spermidine in the PA2-deficient parent and backcross mutant lines, compared with the cultivar (cv.) Birte. Polyamines have been implicated in a wide range of biological processes in plants, including growth, development and stress responses; however the regulation of their metabolism is still unknown (Kumar et al., 1997). The biosynthetic pathways of spermidine synthesis are well-established and involve the synthesis of putrescine via ornithine decarboxylase (ODC) or arginine decarboxylase (ADC) and the conversion of putrescine to spermidine via spermidine synthase (SP), using decarboxylated S-adenosyl methionine (SAM) provided by SAM decarboxylase (SAMDC). Analysis of ADC and SP activities in seeds of PA2-deficient lines provide evidence for a direct biochemical link between PA2 and polyamine metabolism.

Selection and characterisation of PA2 variant lines
A wide screen of sixty-five accessions from the pea (Pisum) germplasm collection led to the selection of JI 1345 as a variant with an elevated total protein content (see below) and a reduced content of the albumin protein, PA2 (Fig. 1A). Genomic blot analyses showed that the mutant line lacked most of the structural genes for PA2 (Fig. 1D) promoter sequence primers that were tested, a promoter region was chosen as a marker, as these primers avoided completely any regions of similarity between PA2 and End1 (see later) genes. Following screening ( Fig. 1A-C), BC mutant lines were selected to grow under controlled environmental conditions for metabolite analysis.
Nitrogen analyses of the RI lines (RILs) were performed on pooled seeds from progeny lines, based on their segregation for le and the PA2 mutation.
Analysis of seeds from RIL pools showed that lines lacking PA2 have an elevated nitrogen content compared with those that contain PA2, in both Le and le backgrounds ( Fig. 2A), suggesting a compensatory increase in other seed proteins in the mutant background. Measurement of the albumin protein fractions from the same RIL pools and parents showed that the PA2 mutation was associated with a significantly lower albumin content (t-test: p < 0.001; Fig. 2B). Gel analyses suggested that the mutant parent, JI 1345, contained a higher amount of pea albumin 1 (PA1b) of approx. M r 6000 than cv. Birte (Fig.   3). Measurements of the yields of this methanol-soluble protein (Fig. 3 a, b) from RI segregants, using a gel scanning and quantification procedure, showed that there was no difference between the PA2 segregant classes but a 21 ± 6 % higher yield of PA1 from the parent JI 1345 compared with cv. Birte ( Fig. 3 d-g). These data suggest that the higher nitrogen content in the PA2 mutant background is not associated with the albumin fraction.

Metabolite profiling of PA2 variant lines
Metabolite profiles were investigated in the mutant line lacking PA2 and in PA2-deficient lines where the PA2 mutation was back-crossed into cv. Birte (BC2) to identify biochemical changes during seed development that may be linked directly or indirectly to the PA2 mutation and to seed nitrogen content.
Since the major part of PA2 accumulation occurs from 20 days after flowering (DAF) onwards (Chandler et al., 1984;Higgins et al., 1987;and   . Consistent with the elevated total seed nitrogen content, the BC and mutant lines lacking PA2 revealed higher total seed protein contents compared with cv. Birte at 20, 30 and 50 DAF (the last represents mature seeds; Fig. 4B).
The protein profiles for immature and mature BC seeds were very similar to those obtained for cv. Birte, apart from the PA2 protein (Fig. 5). In contrast to this, starch content was only slightly elevated in the PA2-deficient mutant, JI 1345, and remained unchanged in the BC (PA2-deficient) line (Fig. 4C).
Metabolite profiles were determined for seeds at 10, 20, 30 and 50 DAF and expressed as described in Roessner et al. (2001a, b). Figure 6 presents a schematic overview of the metabolite changes in both PA2-deficient genotypes relative to cv. Birte, visualized within a metabolic scheme, whereas Figure 1S (Supplement) provides a comprehensive list of all metabolites measured and documents the extent and significance of these changes.
Inspection of the data shows that, at 10-20 DAF, there was a decrease in all the main sugars (sucrose and hexoses) and some minor sugars (such as galactose) in the lines lacking PA2, while glycolytic intermediates (such as hexose-phosphates, 3PGA), most intermediates of the TCA cycle (such as citrate, succinate, malate), several amino acids (such as valine, alanine, lysine and arginine), 4-aminobutyrate (GABA) and polyamines (such as spermidine) were increased. Obviously, a deficiency of PA2 led to increased conversion of incoming sugars to amino acids which, however, was not translated to an increase in protein at this early stage.
At later stages of seed development (30-50 DAF) a different picture emerged.
A lack of PA2 expression led to unchanged or increased levels in major sugars, while intermediates of glycolysis and TCA cycle were unchanged or showed no consistent trend. However, there were clear changes in several individual amino acids governing the increase in total protein synthesis in the PA2 deficient lines. There was an increase in the levels of alanine, phenylalanine, cysteine, arginine and glutamine, while isoleucine, tyrosine and glutamate decreased. Interestingly, lack of PA2 led to a decrease in the level of spermidine, while the level of the precursor of spermidine synthesisarginine -increased. This indicates that a lack of PA2 leads to a specific block in the pathway of spermidine synthesis involving arginine decarboxylase and spermidine synthase. Further analysis of mature seeds using an HPLC technique showed that the decrease in spermidine was accompanied by a similar decrease in spermine levels. The levels of spermine were 18.4 ± 0.7, 15.9 ± 0.8 and 13.7 ± 1.3, and those of spermidine 81.2 ± 2.4, 69.6 ± 3.7 and 59.5 ± 6.9 for Birte, JI 1345 and BC lines, respectively (expressed as nmol gFW -1 , means±SE, n = 3). The values were significantly different from Birte control, according to the Student's t-test (p < 0.05).

Analysis of enzyme activities in the pathway of spermidine synthesis
To investigate the link between PA2 and spermidine synthesis more directly, analyses of the overall activities of arginine decarboxylase (ADC) and spermidine synthase (SP) in developing seeds of PA2-deficient parent and BC mutant lines, in comparison to cv. Birte as a control, were performed. The results show that a PA2 deficiency led to a decrease in the specific activity of ADC, especially at 20 and 30 DAF, where ADC activity decreased markedly compared to the control (Fig. 7A). A deficiency in PA2 also led to a decrease in SP activity at 20 and 30 DAF (approx. 2-fold at 20 DAF), while the activity at 10 DAF was not consistently different in the parent and BC mutant lines in comparison to the cv. Birte control (Fig. 7B). Irrespective of the genotype, there was a general decrease in ADC and SP specific activities during seed development when expressed on a protein basis.

Mapping polyamine biosynthetic and PA2-related genes
The differences observed in polyamine metabolism between the two parent lines (cv. Birte and JI 1345) could reflect biochemical differences as a www.plantphysiol.org on July 6, 2017 -Published by Downloaded from Copyright © 2007 American Society of Plant Biologists. All rights reserved.
consequence of the PA2 mutation or genetic differences associated with linkage to the PA2 mutation. The latter possibility was investigated by isolating and mapping the relevant genes. The genes encoding PA2 were mapped in two populations (JI 281 x JI 399 and JI 15 x JI 1194, using 65-80 progeny lines) to linkage group I, based on polymorphisms in Eco RI digests of genomic DNA (Fig. 1D, e-h). A gene related to PA2, End1, that is expressed predominantly in flowers (data not shown, Gómez et al., 2004), was found to be present in the PA2 mutant line, JI 1345, by Southern analysis (Fig. 1, track l) and by PCR (not shown). It seems likely that End1 represents one of a number of PA2-related genes present in JI 1345 (compare Fig. 1 b, j and l).

DISCUSSION
In the present work, metabolomic approaches have been adopted to investigate the biological function of the albumin protein, PA2 in pea. A mutant lacking PA2 in seeds has been exploited in genetic and metabolite studies.
The fact that PA2 is composed of four hemopexin repeats has suggested www.plantphysiol.org on July 6, 2017 -Published by Downloaded from Copyright © 2007 American Society of Plant Biologists. All rights reserved. possible functions for PA2 as a regulator of metabolic processes (Jenne, 1991). Several mammalian proteins, including matrix metalloproteinases, have hemopexin-like motifs adjacent to their catalytic domains (Le et al., 2007). One such protein, vitronectin, has been shown to interact with serine protease inhibitors and may be important in regulating anti-proteolytic activity during cytolysis (Jenne, 1991). In mammals, hemopexins appear to act as scavengers and transporters of toxic plasma heme, and have been postulated to play a key role in the homeostasis of nitric oxide (Ascenzi et al., 2007).
Hemopexin motifs are likely to be widespread in nature, with prokaryotes having equivalent proteins, predicted to scavenge heme during pathogenicity (Crennell et al., 2000).
Metabolite analyses of the PA2-deficient parent and BC mutant lines showed that a lack of PA2 leads to increased accumulation of seed protein in later stages of seed development (Fig. 4B) which is correlated with changes in the levels of individual amino acids ( Fig. 6; Supplemental Fig. 1S). A lack of PA2 was accompanied by an increase in the levels of alanine, phenylalanine, cysteine, arginine and glutamine. Since accumulation of PA2 contributes approximately 11% to the sulphur-amino acids in pea-seeds (Higgins et al. 1987), the increase in free cysteine could be the direct consequence of decreased PA2 synthesis. The increase in the N-rich amino acids arginine and glutamine and the increase in the glutamine/glutamate ratio indicate that lack of PA2 leads to a higher N status in the seeds. This is consistent with measurements of total N content in seeds, showing an increase in seeds lacking PA2. It is unlikely that the acceleration in seed desiccation observed at 25 and 30 DAF (Fig. 4A) is responsible for the increased N and total protein content in the PA2-deficient lines. Seed protein was also increased when mature seeds were analysed, where no significant changes were observed in seed water content between the different genotypes (data not shown).
In addition to the changes in amino acid pattern and protein synthesis, a lack of PA2 led to decreased amounts of the polyamine spermidine during later stages of seed development (see Fig. 6 in activity of enzymes of spermidine synthesis. First, the decrease in spermidine was accompanied by an increase in arginine, indicating that conversion of arginine to spermidine involving ADC and SP had been inhibited (Fig. 6). Secondly, direct measurements of enzyme activities in developing seeds indicated that inhibition of spermidine synthesis is accompanied by a decrease in the activities of ADC and SP (Fig. 7). It seems unlikely that decreased delivery of decarboxylated SAM as a precursor for the conversion of putrescine to spermidine contributed substantially to the decrease in spermidine accumulation, since the decrease in spermidine was not accompanied by an increase in putrescine levels.
The link between PA2 and polyamines is interesting, since the latter have been implicated in a wide range of biological processes, including growth, It will be interesting to determine why a lack of PA2 leads to a decrease in ADC and SP activities in developing seeds of the mutant and BC lines. Our data show that this is due to biochemical effects rather than genetic linkage.
Although the genes encoding spermidine synthase and PA2 both localised to the pea linkage group I, the two loci were shown not to be closely linked and to have recombined in the generation of the BC lines. We therefore propose that PA2 is biochemically linked to polyamine metabolism in developing seeds, by affecting SP and ADC activities. More studies are required to In conclusion, we identified and characterised a natural pea mutant with a reduced content of PA2 in seeds, offering a number of potential improved seed quality characteristics, including improved digestibility and reduced allergenic potential. Our results show that seeds lacking PA2 have a higher N content and increased levels of overall seed protein, while starch levels were not substantially changed. Metabolic profiling and enzyme measurements revealed a link between PA2 and spermidine accumulation, providing evidence for a role of PA2 in regulating polyamine metabolism in developing seeds.

Plant Material
The pea lines (the cultivar (cv.) Birte and JI 1345, both Pisum sativum L.) single seed descent of 152 progeny lines. Detailed analysis of seed proteins has been carried out at F 5 , F 6 and F 7 generations. One RI line (H86 at F 3 ), having both short internodes and the JI 1345-derived PA2 mutation, was used to introgress the mutation into the cv. Birte. Mutants were identified among the BC1 F 2 segregating population and the back-crossing repeated to give BC2 mutant lines. The pea RI mapping populations, derived from crosses between JI 281 x JI 399 and JI 15 x JI 1194, were used to map the loci encoding PA2, a PA2-related protein, and spermidine synthase.
Plants were grown in a greenhouse with supplementary light and heat in winter months, or under controlled environmental conditions (CEC) for seed development and metabolite analyses. The CEC were: 25°C/day and 20°C/night in a photoperiod 16h day / 8h night at a light intensity of 300 µmol photons m -2 .s -1 . Emerging flowers were tagged, and seed age was expressed in days after flowering (DAF).

Determination of seed weights and water content
Seed development was followed by measurement of seed weight and water content. Calculations of seed fresh weight, dry weight and water content were based on weights determined before and after lyophilisation of seed samples for 24 h. Mature dry seeds were harvested from the lower nodes of the plants (usually up to node 5), avoiding the smaller late pods.

DNA screening methods
Leaves were harvested from young plantlets and used for genomic DNA preparation or leaf imprints were prepared on FTA ® paper (Whatman) and discs removed for PCR analysis, according to the manufacturers' recommendations.
A DNA-based marker was developed to follow the PA2 mutation in RI and back-cross (BC) lines. PA2 genes were isolated from cv. Birte by genome 'walking', using the Advantage ® Genomic PCR and GenomeWalker TM systems www.plantphysiol.org on July 6, 2017 -Published by Downloaded from Copyright © 2007 American Society of Plant Biologists. All rights reserved.
(Clontech), cloned and sequenced. One gene from a Stu I library (PA2stu1) had a promoter of approx. 2.8kb, whereas two from a Dra I library (PA2dra5 and PA2dra8) were shorter with promoter sequences of 200bp (EMBL accession numbers AJ831474, AJ833962 and AJ844650, respectively).
Forward and reverse primers were designed, based on these sequences and tested in various combinations in genomic PCR, using DNA from cv. Birte and JI 1345. The primers PA2-7 (5' AACTTTTGCTAACTATTTACC 3') and PA2-10 (5' CATTTATCCGTCTAACTAAC 3') were selected as the optimal forward and reverse primers, respectively, to screen for PA2 gene variants in RI and back-cross (BC) lines.
The sequence of the PA2-related gene, End1 (EMBL accession number AY324651), was used to design primers based on the promoter region. The primers TB3 (5' AACCAGTGTCCATATATC 3') and TB1_rev (5' AAGGTTATGTTGTGAGC 3') were used to identify variant End1 genes and to screen RI populations. Southern analysis of PA2 and End1 genes was performed, using Eco RI digests of genomic DNA.

Protein screening methods
Mature seeds were screened, using meal that was removed with a sharp scalpel blade from seeds before sowing. Immature seeds were ground to a fine meal, following lyophilisation. Proteins were extracted from seed meal either by heating directly in 5x LDS buffer or by shaking in 20mM ammonium acetate, pH 5.0 (25 mg seed meal per ml) for 1.5h at 4°C. Ammonium acetate supernatants (albumin fractions) were analysed on polyacrylamide gels (NuPage Bis-Tris 12% gels in MOPs buffer, Invitrogen), following addition of 5xLDS buffer and DDT to 0.05M, and incubation at 70°C for 10 minutes.
Albumins were fractionated according to methanol solubility, following addition of four volumes of 75% methanol and precipitation of insoluble proteins at -80°C for 30 minutes. Soluble proteins were recovered from supernatants following addition of four volumes of acetone at -20°C. Samples were analysed on polyacrylamide gels (as above, but using 4-12% gradient gels in MES buffer) and the relative amounts of methanol-soluble proteins determined by scanning the gels. Protein concentrations were determined using a BioRad protein assay kit and bovine serum albumin in an appropriate buffer as standard.
Total seed nitrogen was determined for meals derived from cotyledons, following removal of testas. The relative contents of total carbon and nitrogen in dried, powdered meal samples were measured using an elemental analyser (Vario EL; Elementaranalysensysteme, Hanau, Germany). A minimum of six seeds was analysed per sample for parent lines; where RILs were pooled according to genotype, the minimum pool size was 46 seeds, with at least two seeds per line. N x 5.44 was used as the nitrogen to protein conversion factor for pea (Mossé, 1990). All measurements were performed on triplicate samples.

Metabolite analyses
Pods were frozen in liquid nitrogen and seeds were manually separated from the pods under liquid N 2 before being ground to a fine powder using a liquid www.plantphysiol.org on July 6, 2017 -Published by Downloaded from Copyright © 2007 American Society of Plant Biologists. All rights reserved. nitrogen mill (Retsch Schwingmühle M200, Haan, Germany). Ground seeds (50 mg) were extracted according to a method modified from that described previously (Fait et al., 2006) and the relative metabolite contents were determined by gas chromatography-mass spectroscopy (GC-MS) as previously described and validated (Roessner-Tunali et al., 2003), using ribitol as internal standard (Roessner et al., 2000). The chromatograms and mass spectra were evaluated using the MassLab program (ThermoQuest). A retention time and mass spectral library for automatic peak quantification of metabolite derivatives was implemented within the MassLab method. Data are normalized to the mean response calculated for the cv. Birte samples as described by Roessner et al. (2001a).
Free soluble spermine and spermidine were also determined using HPLC (Dionex GmbH, Idstein, Germany) after dansylation and quantification with a fluorescence detector, according to Smith and Davies (1985).
Protein content was measured according to Eastmond and Rawsthorne (2000). The protein concentration was determined using the dye-binding assay (Bradford et al., 1976) with bovine serum albumin as the standard.
Starch content was measured as described in Geigenberger et al. (1998).

Analysis of enzyme activities
Enzyme extracts were prepared from frozen seed material according to Geigenberger and Stitt (1993). Arginine decarboxylase (EC 4.1.1.19) was measured as in Pérez-Amador and Carbonell (1995). Spermidine synthase activity (EC 2.5.1.22) was measured by monitoring the production of spermidine from putrescine and DcSAM by HPLC. Assays were incubated at 37°C. The reaction mixture contained 0.1 M Tris-HCl (pH 8.0) buffer, 0.1 mM putrescine, 80 µM DcSAM and an aliquot of protein extract. After 30 min, the reaction was stopped by heating (5 min, 95°C). In the control assay, the reaction mixture was placed immediately at 95°C. Free soluble spermidine was quantified with a fluorescence detector via HPLC (Dionex GmbH, Idstein, Germany) after dansylation according to Smith and Davies (1985). The enzyme assays were diluted 1:1 with equal amounts of 0.2 N HClO 4 and 1.5 M Na 2 CO 3 before dansylation. 20 µl of sample were injected onto a reverse phase Supelcosil TM LC-18 column (Supelco, Bellefonte, USA) protected by a guard column. Samples were eluted from the column with a solvent gradient (v/v) of water:methanol, changing from 70% to 100% in 15 min at a flow rate of 1 ml/min. The column was washed with 100% methanol for 5 min and re-equilibrated at 70% methanol for 5 min before the next sample was injected. Eluates from the column were detected by an        Blots were washed at medium (50 0 C, a-j) or high (65 0 C, k, l) stringency. Comparison of total seed protein content (N x 5.44) and albumin content in parent lines ed) and RIL segregant classes (blue). RILs have been pooled, according to the segregation of