A Suite of Lotus japonicus Starch Mutants Reveals both Conserved and Novel Features of Starch Metabolism

The metabolism of starch is of central importance for many aspects of plant growth and development. Information on leaf starch metabolism other than in Arabidopsis is scarce. Furthermore, its importance in several agronomically important traits exemplified by legumes remains to be investigated. To address this issue, we have provided detailed information on the genes involved in starch metabolism in Lotus japonicus and have characterised a comprehensive collection of forward and TILLING reverse genetics mutants affecting five enzymes of starch synthesis and two enzymes of starch degradation. The mutants provide new insights into the structure-function relationships of ADPglucose pyrophosphorylase and glucan, water dikinase1 in particular. Analyses of the mutant phenotypes indicate that the pathways of leaf starch metabolism in L. japonicus and Arabidopsis are largely conserved. However, the importance of these pathways for plant growth and development differs substantially between the two species. Whereas essentially starchless Arabidopsis plants lacking plastidial phosphoglucomutase grow slowly relative to wild-type plants, the equivalent mutant of L. japonicus grows normally even in a 12 h photoperiod. In contrast, the loss of GWD1, required for starch degradation, has a far greater effect on plant growth and fertility in L. japonicus than in Arabidopsis. Moreover, we have also identified several mutants likely to be affected in new components or regulators of the pathways of starch metabolism. This suite of mutants provides a substantial new resource for further investigations of the partitioning of carbon and its importance for symbiotic nitrogen fixation, legume seed development, and perenniality and vegetative re-growth.


8
We were able to identify L. japonicus genes encoding all of the classes of enzymes involved in starch metabolism in Arabidopsis, but there were several differences in isoform representation. For instance, we could not find L. japonicus sequences orthologous to Arabidopsis genes encoding the glucan, water dikinase GWD2 and the β -amylases BAM2 and BAM4 (Supplemental Table S2). These genes may be present, but in an unsequenced region of the genome (9% of the gene space has not been sequenced; Sato et al., 2008), or expressed at a low level and so not represented in the EST collection. Alternatively, there may be differences in the composition of some gene families between the two species. We also found duplications of several starch metabolism genes in L. japonicus, including AGPase large subunit isoform 2 (APL2), starch synthase 2 (SS2), granule-bound starch synthase (GBSS), α amylase 3 (AMY3), β -amylase 3 (BAM3), and cytosolic glucan phosphorylase (PHS2, also known as Pho2 in some species). Duplication of genes encoding SS2 and GBSS exist in numerous species, including other legumes and cereals (Pan et al., 2009). Interestingly, the duplications appear to have independent origins in different groups of plants. The legume duplication resulted from a whole-genome duplication within Rosid clade I, whereas the cereal duplication resulted from the whole-genome duplication at the base of the grass lineages, well after the divergence of the monocots and dicots (Pan et al., 2009).

Forward Genetic Screens to Identify Mutants Affected in Starch Metabolism
To identify mutants, we used a collection of lines impaired in starch metabolism from a screen on 1428 M2 families (17,100 plants) from seeds of Gifu mutagenized with EMS (Perry et al., 2003). Our screen utilized the fact that decolorized leaves of WT plants stain strongly with iodine solution at the end of the day when starch content is high, and much less strongly at the end of the night when starch content is low (Fig. 3). We used this to isolate mutants that had either lower levels of leaf starch than WT plants at the end of the day ('synthesis mutants') or higher levels of leaf starch than WT plants at the end of a dark period ('degradation' mutants). Because starch content at the end of a normal night varied from batch to batch and with glasshouse conditions, screening for degradation mutants was usually performed on plants subjected to an extended night of up to 44 h. The clearest results were obtained on young plants (about 4 weeks old) in which starch content is generally low at the end of the night (Fig. 1).
The screens led to the selection of ten synthesis and six degradation mutants (Fig. 3).
Allelism tests suggested that the mutations defined at least five loci for the synthesis mutants, and four loci for the degradation mutants. The selected mutant lines were outcrossed at least once to MG-20, to reduce numbers of background mutations introduced by EMS mutagenesis and to establish a mapping population. Segregation ratios of the F2 progeny (data not shown) indicated a recessive monogenic mode of inheritance for all the starch mutant phenotypes, consistent with the nature of such mutations in other species.

Mutations Affecting Starch Metabolism
To discover genes necessary for starch metabolism in L. japonicus, we took two approaches: mapping, to identify the mutations responsible for the phenotypes identified above; TILLING to identify mutations in selected genes encoding enzymes known to be necessary for starch metabolism in other plant species (Supplemental Table S2). TILLING was also used to identify additional mutant alleles of genes for which only one mutant allele was isolated from the forward genetic screen.  Table S3). The mapping interval thus identified (Supplemental Table S4) was confirmed by further mapping on the mutant individuals.
Using information from the genome sequence of L. japonicus (Sato et al., 2008), we searched within the confirmed interval for candidate genes (see Supplemental Table S2; Fig. 2). Synteny with the genomes of the legume species soybean (Glycine max; http://www.phytozome.net/soybean) and Medicago truncatula (http://www.medicago.org/genome/) were also used in some cases where the L. japonicus genome sequence was not available. Where sequencing of candidate genes (for primers see Supplemental Table S5) revealed mutations, biochemical and genetic approaches were used to check rigorously that the mutations were indeed responsible for the starch phenotype. As described below, we were able to identify the mutated genes in six out of the ten synthesis mutants, and three out of the six degradation mutants.
TILLING was carried out using two DNA populations of L. japonicus (GENPOP and STARPOP;Perry et al. 2009 Table S5).

Mutations in the Starch Synthesis Genes LjPGI1, LjPGM1, LjAPL1, LjAPL2 and LjAPS1
Based on knowledge gained from Arabidopsis (see Introduction and Supplemental Fig. S1), we examined whether the almost starchless and low-starch plants (synthesis mutants) selected in the forward screen lacked pPGM, pPGI, or AGPase. As described below, mutations affecting these three enzymes accounted for the phenotypes of six of the ten synthesis mutants.
Three of the synthesis mutants from the forward screen (SL4308-12, SL4715-2 and SL5069-2) had leaf starch contents that were only 10% of WT values (Table I), but their embryo and root starch contents were indistinguishable by iodine staining from those of WT plants ( Fig. 3 for SL4308-12, and data not shown). This pattern of starch distribution is also seen in the Arabidopsis pgi1 mutant (Yu et al., 2000). The three mutants were shown by crossing to be allelic, and the mutations in all three mapped to the same interval at the top of chromosome I. The LjPGI1 gene is located in this interval ( Fig. 2; Supplemental Tables S2 and S4). Sequencing of the PGI1 gene in the three mutants revealed two mutations that create stop codons (pgi1-1, SL4715-2; pgi1-2, SL4308-12) and one (pgi1-3, SL5069-2) at a splicesite junction ( Fig. 4 and Supplemental Table S6). Native gel electrophoresis followed by activity staining on protein extracts from leaves revealed two bands of PGI activity in WT extracts, attributable to the cytosolic and plastidial isoforms of the enzyme (Shaw et al., 1970). The band attributable to the plastidial isoform was missing in extracts of the mutants (Supplemental Fig. S2), consistent with the specific loss of the plastidial isoform of PGI in the pgi1 mutants.
One of the synthesis mutants, SL4725-4, appeared from iodine staining and starch quantification to lack starch in leaves, stems, roots, and embryos, suggesting that it might be a pgm1 mutant ( Fig. 3; Table I). Arabidopsis, tobacco and pea mutants lacking pPGM are starchless in all plant parts examined (Caspar et al., 1985;Hanson and McHale, 1988;Harrison et al., 1998). In a second mutant (SL4867-11) with a strongly reduced starch content relative to WT plants in all organs examined, the mutation mapped to an interval of about 10 centi-Morgan (cM) on chromosome V, in which the PGM1 gene is located ( Fig. 2; Supplemental Tables S2 and S4). Sequencing revealed that the PGM1 gene in SL4725-4 contained a mutation affecting a splice-site junction, and the gene in SL4867-11 contained a mutation predicted to result in the amino acid change G95D. These mutants are referred to as pgm1-4 and pgm1-5 respectively ( Fig. 4; Supplemental Table S6). Native gel electrophoresis followed by activity staining on protein extracts from leaves revealed two bands of PGM activity in WT extracts, attributable to the cytosolic (PGM2) and plastidial (PGM1) isoforms of the enzyme (Harrison et al., 1998). The band attributable to the plastidial isoform was missing from extracts of the mutants, confirming that chloroplastic PGM activity is strongly reduced or absent in both cases (Supplemental Fig. S2). We used TILLING to select lines homozygous for three additional mutant alleles of the PGM1 gene. Two of the alleles (pgm1-2, from SL755-1 and pgm1-3, from SL1837-1; Fig. 4; Supplemental Table S6) contained mutations creating stop codons; both of the mutant lines appeared from iodine staining to lack starch in all plant parts and were identical in phenotype to pgm1-4 ( Fig. 3; described in more detail below). The mutation in the third allele (pgm1-1, SL4490-1) was predicted to result in the amino acid change D436N. Activity of the plastidial isoform of PGM was reduced in pgm1-1 (Supplemental Fig. S2), but no reduction in starch content of the leaves was apparent from iodine staining (Fig. 3).
We used similar approaches to identify the mutation accounting for the phenotype of another synthesis mutant, SL5127-5, in which leaf starch content at the end of the day is typically reduced by about 80% (Table I) Table S2). In Arabidopsis, loss of the small subunit (in the adg1 mutant) almost eliminates AGPase activity and starch synthesis (Lin et al. 1988a;Wang et al., 1998a), and loss of the major leaf isoform of the large subunit (in the adg2 mutant) reduces activity by 95% and starch synthesis by 60% (Lin et al., 1988b;Wang et al., 1997). Sequencing revealed a single mutation in the APL1 gene of SL5127-5, predicted to result in the amino acid change S400L. To discover whether this mutation can account for the starch phenotype of the mutant, we assayed AGPase activity in extracts of WT and mutant leaves. Activity was 93% lower in mutant than in WT extracts (Table II). Activity in extracts made from mixtures of WT and mutant leaves was 98.5% of that predicted from separate extracts of the two genotypes, hence the large difference in activity is likely to result from loss of APL function in the mutant rather than the presence of inhibitory substances in the mutant leaf.
To understand further the importance of AGPase subunits in L. japonicus, we used TILLING to identify mutations in a second large subunit gene, APL2, and in the single small subunit gene, APS1. For APL2, we isolated four alleles containing mutations predicted to have a deleterious effect on activity of the encoded protein. None of these had an effect on starch content (assessed by iodine staining) in any of the organs examined, including roots, leaves, and embryos (data not shown). In contrast, plants carrying one of the four mutant alleles identified for APS1 (aps1-1, from SL529-1), which had a mutation predicted to bring about the amino acid change A111T (Supplemental Fig. S4; Supplemental Table S6), displayed a strong reduction in starch content in leaves, roots, and embryos (Fig. 3). AGPase activity in leaves of this mutant was about 70% lower than in wild-type leaves (Table II), and leaf starch content was reduced by up to 90% (Table I).

Mutations in the Starch Degradation Genes LjGWD1 and LjGWD3
Following the same approach as for the starch synthesis mutants, we identified the mutations responsible for the starch-excess phenotype of three of the starch degradation mutants. The mutations in SL5215-2 and SL5358-3 mapped to the same interval on chromosome IV, and the mutation in line SL5104-12 mapped to the top of chromosome V  Table S2). In Arabidopsis, the gwd1 (aka sex1) mutant has a severe starch-excess phenotype, and reduced growth under short-day conditions (Caspar et al., 1991;Yu et al., 2001). The gwd3 (pwd) mutant (Baunsgaard et al., 2005;Kötting et al., 2005) also has a starch-excess phenotype, although less severe than that of gwd1. Sequencing revealed that SL5215-2 and SL5358-3 both carry a mutation in GWD1, while SL5104-12 has a mutation in GWD3. Although SL5215-2 and SL5358-3 were independently selected, they carry the same mutation in GWD1, predicted to result in the amino acid change E566K ( Fig.   4; Supplemental Table S6). Plants carrying this mutation are referred to as gwd1-1 mutants.
The mutation in GWD3 affects a splice-site junction and results in the nucleotide change G7871A ( Fig. 4; Supplemental Table S6).
To confirm that the mutations identified in these two genes were responsible for the starch excess phenotype of the gwd1-1 and gwd3-1 mutants, we generated additional mutant alleles by TILLING. We targeted a region of the LjGWD1 gene encompassing the sequence   Table S6) had a clear starch excess phenotype in leaves (Fig. 3). The mutation in this line is predicted to affect a residue (amino acid change G980E) that is identical across all the predicted protein homologs of LjGWD3 we analyzed, and lies within a conserved motif (PWD_ARATH, Q6ZY51; PWD_ORYSA, NP_001066613; putative homologs in Vitis vinifera, XP_002265211; Ricinus communis, XP_002518612, and Sorghum bicolor, XP_002453659; multiple sequence alignment not shown). The starch-excess phenotypes of both gwd3-1 and gwd3-4 strongly suggest that LjGWD3 plays an important role in the degradation of starch in leaves of L.
japonicus. It seems likely that, as in Arabidopsis, LjGWD3 acts in synergy with LjGWD1 to interacts with other amino acids in the β -helix domain in which it is located. These interactions are lost when a leucine is substituted for this serine residue. We suggest that the serine S400 may be important for the 3-D structure of the large subunit itself, and hence for heterotetramer stability as a whole.
The mutation in gwd1-1 also provides new information about the structure-function

Importance of Starch Metabolism for Plant Growth, Development, and Nodule Function in L. japonicus
We used our synthesis and degradation mutants to examine whether loss of normal starch metabolism affected growth and development, and especially the legume-specific capacity for nitrogen fixation in Rhizobium-containing root nodules. The essentially starchless phenotype of the pgm1-4 mutant ( Fig. 3; Fig. 5 Most gwd1 flowers did not produce pods, and when they did, pods and embryos often aborted early in development. However, fully developed mutant pods and seeds could be obtained by detaching flowers and placing the peduncles in MS liquid medium supplemented with 3% sucrose (not shown). This result suggested that the effect of the mutation on the production of viable seed was maternal rather than due to a defect in embryo development. As a further test, we analyzed embryos and seeds from 40 pods each of WT and heterozygous (GWD1-2/gwd1-2) plants from the same segregating population. Both the number of seeds per pod and the average seed weight were very similar for the two genotypes (an average of 9.1 and 10.2 seeds per pod for heterozygous and WT respectively, and an average seed weight of 1.0 mg for both genotype). Similar results were obtained for the gwd1-3 mutation (not shown). These data support the idea that reduced production of viable seeds on homozygous mutant plants is a maternal effect. However, genotyping of each individual mature embryo from heterozygous plants was not wholly consistent with a maternal effect. Out of 63 mature seeds, only seven contained homozygous mutant embryos (one in nine rather than the expected one in four ratio). Thus, mutant embryos were disadvantaged whether in mutant or heterozygous pods. The effect of the gwd1 mutations on fertility thus appears to be complex.
Gametophytic, maternal, and embryo effects may all be involved. In general, our mutants show that the phosphorylation of starch via the glucan, water dikinases GWD1 and GWD3 is essential for normal starch degradation in most parts of the L. japonicus plant and that this, in turn, is crucial for normal plant growth and development.
Legume nodules generally contain starch in cortical cells and non-infected cells within the infected zone (Gordon, 1992;Gordon and James, 1997;Szczyglowski et al., 1998). In pgm1-4 mutants inoculated with Mesorhizobium loti, nodules had a normal pink color, and their size and number per root system were similar to those of WT segregants (Fig. 5). In spite of the apparent lack of starch in the nodules, assays for nitrogen fixation (acetylene reduction) revealed no statistically significant differences (Student's t-test p-value of 0.05) between mutant, heterozygote and WT plants 36 days after inoculation (0.57 ± 0.06, 0.75 ± 0.11, and 0.67 ± 0.20 μ mol ethylene h -1 root system -1 , respectively; means ± SE of 5 root systems). wild-type and heterozygous plants. Both batches showed that the gwd1 mutants were capable of nitrogen fixation (Supplemental Fig. S3D). Taken together with the results of measurements on the nodules of pgm1-4 mutant plants, these data indicate that starch storage and normal starch metabolism are not essential for nodule function.

CONCLUSIONS
We have linked the identification of candidate genes from rough mapping with TILLING reverse genetics to assemble a comprehensive suite of starch mutants in the model legume, L.
japonicus. This is a valuable new resource with which to examine the partitioning of carbon in relation to several crop traits of agronomic importance such as nitrogen fixation and perenniality. Our research has additional significance. It shows that the main components of the pathways of leaf starch metabolism are conserved across different families and life histories. In contrast, the control of flux through these pathways and their importance in sustaining normal plant growth are very different in Arabidopsis and L. japonicus. Whereas a severe deficiency in starch synthesis has much less effect in L. japonicus than in Arabidopsis, loss of the capacity for starch degradation has much more profound consequences for plant growth. Further investigation of these differences and the genes affecting starch metabolism that were previously unknown could lead to a better understanding of the regulatory mechanisms linking the metabolism of starch with plant growth. Our analysis also revealed significant differences in starch metabolism between plants of MG20 and Gifu, two WT ecotypes of L. japonicus with large difference in several other traits, including biomass production. This finding is particularly interesting in the light of the recent demonstration that rates of starch turnover are positively, causally linked to productivity in Arabidopsis (Sulpice et al., 2009). We suggest that L. japonicus is an excellent system with which to test the wider applicability of this finding.

Plant Material and Growth Conditions
Initial screening for starch metabolism mutants was carried out as described by Perry et al. (2003;2009  For in vitro culture of pods, peduncles carrying mature flowers were cut and the ends immediately immersed in distilled water. Peduncles were then surface sterilized with 10% (v/v) bleach (1% available chlorine), and rinsed repeatedly in sterile water. After re-cutting, peduncles were inserted through a hole in the lid of a 50 mL tube of sterile Murashige and Skoog medium (micro and macro elements including vitamins, pH 5.7) with 3% (w/v) sucrose, then kept for 4 weeks in a growth chamber with 16-h light, 8-h dark.
Upon request, all novel materials described in this publication will be made available in a timely manner for non-commercial research purposes, subject to the requisite permission from any third-party owners of all or parts of the material.

In-silico Analysis of Gene and Protein Sequences
Genes encoding known enzymes of starch metabolism in L. japonicus were identified from

Iodine Staining and Starch Quantification
Tissue was heated in 80% (v/v) aq. ethanol to remove pigments before staining with iodine (Lugol's solution), except for embryos, which were incubated in chloroform:ethanol:water  Tables   Table S1. Diurnal changes in leaf starch content through development of L. japonicus plants.    Starch content (mg Glc equivalents g -1 fresh weight)