Figure 1A shows a phylogenetic tree with clear separation of the plant from the bacterial sequences. This is in agreement with earlier reports regarding cyanobacterial and plant SPS sequences (Langenkamper et al.,2002). Figure 1B shows a parsimony tree of plant sucrose synthase sequences and their separation into at least three sub - families. These three families have been labeled A, B and C. Of the 71 plant sequences, 20 monocot sequences clustered within sub-family A, 35 dicot sequences within a large sub-family B and 16 dicot and monocot sequences within sub-family C. All three sub-families are characterized by the presence of paralogues within a single species. Early classification of sucrose synthase indicated 4 main classes, class I consisting entirely of enzymes from monocots, class II, III and IV consisting entirely of enzymes from dicots (Komatsu et al., 2002, Sturm et al., 1999). Studies on SPS sequences reported clustering of SPS sequences into at least three families and every plant had a representative of each family in their genome. The clear distinction between dicot and monocot proteins is seen in our study as well; however dicot derived sequences are distributed into at least two subfamilies, one of which has monocot representatives. The phylogenetic trees generated using both Neighbour Join and Parsimony methods were in agreement and substantiated this clustering. The separation of A, B and C subfamilies is supported by bootstrap values of 88, 97 and 100 respectively. Within the subfamilies A and C, bootstrap support for all nodes range from 70-100 across the family, whereas within the dicot sub-Family B, there appear to be further divisions separating the family into at least 3 groups with bootstrap support of 77, 100 and 62. The clustering also appears to correlate with taxonomic separation, with representatives of the order Caryophyllales (Dianthus, Chenopodium and Beta vulgaris) clustering away from the Asterids sequences (Coffea, Nicotiana and Solanum) which, in turn, are separated from all the other representatives of the Rosids subclass (that includes Arabidopsis, Eucalyptus, Gossypium, Populus, Citrullus, Citrus and legume representatives). The dicot species that are represented in both Families B and C are Arabidopsis, Citrus, Pisum, Beta, Coffea and Solanum sp. The Arabidopsis representatives include six SUS proteins – SUS1 to SUS6 (Baud et al., 2004). SUS1 and SUS4 cluster with Family B, while SUS2, SUS3, SUS5 and SUS6 in Family C. There are some dicot sequences that do not have a homolog in B but are represented in sub-family C. These include a sequence each from Pyrus pyrifolia and Pyrus communis besides two homologs from Craterostigma plantagineum. Within the dicot family B, there is ample evidence of duplication and divergence. This feature is exemplified by these pairs- Medicago truncatula and M. sativa, Phaseolus aureus and P.vulgaris; Coffea arabica and C.canephora; N.langsdorfii and N.tabacum. This feature is seen in the reconciled tree (Figure. 1D).Table1

In Family C with 16 representative sequences, there are 4 homologs from A. thaliana (SUS2, SUS3, SUS5 and SUS6; SUS5 and SUS6 cluster separately), two from C.plantagineum, one each from B.vulgaris, S.tuberosum, C.arabica and C.unshui. Even here, duplication and divergence events are evident. Homologs from Orysa and Maize are also seen in Family C along with the other dicot sequences. The Orysa SUS sequence has between 55 - 65% identity with all the four Arabidopsis sequence representatives in this family. The ATSUS5 and ATSUS6 proteins cluster separately within this family, as has been reported earlier (Baud et al., 2004).

Most sequences in sub-family A cluster with their closest plant relative. This family is represented by two wheat homologs, with 77% sequence identity (Wheat Q42223 and Wheat 082073); Maize Sus1 and Sus2 with 80% identity; SUS1 and SUS2 sequences from Tulipa gesneriana with 85% identity; SUS1 and SUS2 sequences from H.vulgare with 78% sequence identity. Sub-family B also has two homologs from B.oldhamii with 81% identity; that these sequences arise out of duplication events is shown in Figure 1C. The Oryza sativa SUS2 and SUS3 proteins are nearly 90% identical. There are two homologs from P.distinctus that are both represented in Family B. Other Family B representative sequences include M. acuminata, O. goldiana and Lolium perenne. The sequences represented in subfamily A differ from their homologs in subfamilies B and C in the presence of two leucine zipper regions in many representatives of this family (LNDRIRSLDALQAALRKAEEHL, and LSAHTNELVAVFSRLNQGKGML) (Figure 2). Three monocot sequences in Family C have a variant of one of the bzip motifs (LTDNKKPLLDYLLALSHRGDKL in O. sativa and Maize species). This finding is consistent with reports on the role of a helix loop helix transcription factor in the regulation of carbohydrate metabolism in response to sugar levels in the human liver, and in the sucrose induced repression of translation (SIRT) response in plants (Weise et al., 2004). Sub-family B sequences are mostly characterized by four motifs for Asn-glycosylation (NYSD/ NHTD), protein kinase C and casein kinase 2 phosphorylation sites and the MYRISTL pattern. The presence of additional cAMP phosporylation site motif (KKLS/KRLT/RRLT), ATP-GTP-A or leucine zipper motifs characterize the difference within the sub-groups in sub-family B representatives. Most Family C representatives have an amidation motif (GGKK/NGRR/QGKR) that is absent in the other two families. Features shared amongst all representatives of the three sub-families include Asn-glycosylation motifs (NYSD), PKC phosphorylation site patterns (SLR, TEK, SAK in subfamily B, SSK and TPK in Family A), casein kinase 2 phosphorylation patterns(SRFD, TYQE in sub-family C and TFQE in sub-families A and B) and MYRISTL (GQYESH, GVTQCT, GTEHTH, GNLVAS in B and C, with GLVAC/T in subfamily A). We note a feature of the MYRISTL motif (N-myristoylation site) in that some motifs are dicot sequence-specific (GILQS/HH, GL/YPDTG) and differ between representatives of subfamilies B and C in a single amino acid. N-myristoylation is a lipid modification ensuring the proper function and intracellular trafficking of proteins involved in many signaling pathways. The presence of this feature suggests multiple roles of this enzyme and multiple localization of the protein helping them reside in more than one compartment of the cell. Attachment of the myristoyl residue provides hydrophobicity that influences the partitioning of proteins to cellular membranes and can serve to promote protein-protein interactions (Johnson et al., 1994). It has been shown that phosphorylation of sucrose synthase decreases its hydrophobicity and releases the protein from the membrane while dephosphorylation increases association with membranes (Winter et al., 1998). The existence of multiple phosphorylation, N-glycosylation and leucine zipper motifs has been reported in Maize SUS3 (Carlson et al., 2002). From this prediction, it appears that all plant sucrose synthase sequences have some or all of these patterns with variants that could be phylogenetic cluster specific. The sequence variations in sucrose synthase proteins were then examined from a structural perspective. A representative alignment of the sequences derived from all three subfamilies was further examined using ConSurf (http://consurf.tau.ac.il). This structure based sequence conservation analysis was made using the sucrose phosphate synthase (SPS, PDB 2r60) as a template. This is shown in Fig.3. As is evident from this figure, the variable regions are located in the N terminal region (7-8, 26-27 and 68-73) and a few sectors in the C terminus (398-401, 423-427, 447-448). The most variable segments comprise those between 239-253, 258-262 and 306-309. Although the functional significance of these variations is unclear (apart from a generalized assumption that these regions are not essential for function), it is possible that some of these segments might modulate the activities of these enzymes (allostery) or might be involved in interaction with other co-factors in the cellular mileau.

It has been suggested that the distribution of SUS in organelles not involved in sucrose metabolism may indicate functional roles beyond sucrose degradation (Gerrits et al., 2001; Oswald et al., 2001; Subbaiah et al., 2006). A related finding is that many sucrose sythase sequences show a mitochondrial targeting peptide (mTP). This study was carried out on maize sucrose synthase proteins (Subbaiah et al., 2006). Many of these polypeptides are synthesized in the cytosol and subsequently translocated to other organelles by virtue of a targeting sequence. The TargetP server was used to scan for these signal sequences. Based on the location of these target peptides at the N terminus of these proteins, it is seen that many more of sub-family A monocot representatives have a mitochondrial localization relative to subfamily B or C sequences. Only 3 (8.5%) of subfamily B and 2 (12.5%) of subfamily C sequences were predicted to have the mTP peptide relative to 12 (63%) of subfamily A sequences (Table 2). The scores for each target peptide sequence were generated with specificity of >90. This sequence analysis suggests that one protein each from subfamilies B and C is likely to be involved in secretory pathways. While not much is known on the role of sucrose synthase in the mitochondria, it is suggested that its role may be similar to that of hexokinase interaction with mitochondrial VDAC (voltage dependent anion channel). This interaction regulates the opening of the mitochondrial permeability transition pore and also affects nutrient or sugar signaling and apoptosis (Azoulay-Zohar, 2003). The analysis of the SUS proteins in this dataset suggests that the presence of a targeting peptide is more of a monocot SUS sequence specific feature and may not apply across all plant sucrose synthase proteins.

The multiple sequence alignment of the protein sequences was carried out using the consistency based method of PROMALS (PROfile Multiple Alignment with predicted Local Structure (http://prodata.swmed.edu/promals/promals.php). Sequence alignment with PROMALS is a two step process. In the first step, highly similar sequences are progressively aligned with a weighted sum-of-pairs measure of BLOSUM62 scores resulting in a set of sequences or pre-aligned groups that are relatively divergent from each other. In the second alignment stage, one representative sequence is selected from each pre-aligned group, and PSI-BLAST is used to search for homologs from the sequence database UNIREF90. The profiles generated here are derived from the PSI-BLAST alignments and PSIPRED secondary structure prediction. A matrix of posterior probabilities of matches between positions is obtained by forward and backward algorithms of a profile-profile hidden Markov model. These matrices are used to calculate a probabilistic consistency score and the representative sequences are further aligned according to a consistencybased scoring function. The pre-aligned groups obtained in the first stage are subsequently merged with the multiple alignments of the representatives. The accuracy of this protocol derives from the combined use of homologs from a structure database and secondary structure predictions using PSIPRED. While several available tools were used for MSA generation, the representative sequences derived from the PROMALS output were used to search the stress sequence database for homologs. The secondary structure predictions of sequences from the three families show minor differences in the N terminal regions containing the putative target peptide region (Fig.4). While sucrose metabolism proteins are generally believed to be cytosolic proteins, predicted trans-membrane segments were seen in programs such as AveHas as well as TMpred (http://www.ch.embnet.org/software/TMPRED_form.html). Figure 4 shows the results on the amphipathicity, hydrophobicity and similarity of sequences in each of the three subfamilies generated using the AveHas programme. Sub-family A is characterized by the presence of 11 amphipathic peaks (A1-A11) with the angle set at 180° (as is appropriate for a β-strand) and 6 hydropathic peaks (H1-H6). Sub-family B is characterized by 9 amphipathic peaks (A1-A9) and 7 hydropathic peaks (H1-H7). Sub-family C is characterized by 8 amphipathic peaks (A1-A8) and 7 hydropathic peaks (H1-H7). In all these three plots, the average similarity plot follows the hydrophobicity plot and the major difference between the three families is in the N-terminal region- specifically between residues 7-90 in the alignment. We also note differences in the hydrophobic character of this region between families, with increased hydrophobicity in sub-family C while sequences from sub-family A (monocots) are more amphipathic in nature. The most conserved region in the alignment of sub-family A representatives (150 – 445 residues) is characterized by fewer hydrophobic amino acid stretches (A5-A10 and H2-H4). The most conserved region of the alignment in sub-family B (291-443) corresponds to H2-H5 and there is complete conservation in hydrophobic residues within this region in all the representatives. Predicted trans-membrane helices also lie within this region (249-480). In sub-family C, there is likelihood of trans-membrane segments between position 20-80 of the aligned sequence besides one between the 270-295 positions of this alignment. Thus, within each of sucrose synthase family clusters, there appear to be structural differences that are not understood, but potentially contribute to the phylogenetic separation of the families and potential differences in their functional role in abiotic stress.

The sequences identified using PROMALS represent sequences across all 3 subfamilies. There were 6 representatives from sub-family A, 8 from sub-family B and 11 from sub-family C. These sequences were searched against the EST databases for nine crops, five monocots (barley, sorghum, rice, maize and wheat) and 4 dicot species (Arabidopsis, Medicago, Soybean and Tomato). The reference sequences were also searched against the EST datasets derived from stress cDNA libraries from each of the nine crops. The percentage of identified hits to reference sequences from the three subfamilies in the EST database (not derived from abiotic stress) ranged from 0.008% (AtSus6, sub-family C against Rice EST database accessed at JCVI) to 0.04% (Eucalyptus grandis, sub-family B against the Maize database accessed at JCVI) with an average of 0.026%. The percentage of identified hits to reference sequences from the three families in the stress EST datasets for the nine crops ranges from 0.005% (all reference sequences against the Arabidopsis ESTs derived from stress cDNA libraries) to 0.44% (all reference sequence against the Tomato EST dataset derived from stress cDNA libraries) with an average of 0.08%. On the whole, 24 reference sequences when searched against stress cDNA libraries returned a larger number of hits than when searched against normal EST datasets that do not include sequences derived from stress libraries, after normalizing the number of hits for the size of datasets. The only exception was the Arabidopsis stress EST dataset with very little representation of sucrose synthase homologs. There was no significant difference between the number of homologs to reference sequences having both domains (sucrose synthase and glycosyl tranferase1) as compared to a reference sequence with only one of the two domains. Interestingly, homologs to AtSus6, OsA and DaucaSus2 were found to have a significantly greater representation than the other reference sequences in the stress derived EST datasets (5% significance using the Z-test) across six of the nine crops for which stress EST datasets were screened. The exceptions were Arabidopsis, Maize and Rice datasets where the percentage hits were higher in non-stress than stress datasets. AtSus6 and OsA belong to sub-family C and, in terms of sequence features, they are characterized by the bzip + amidation patterns, have transmembrane predictions and lack a target peptide region. Dauca Sus2 from sub-family B has a mitochondrial target peptide but lacks the bzip and amidation pattern. In the Arabidopsis stress EST data collection, the paucity of transcripts to sucrose synthase could be because of the tissues from which the cDNA libraries were derived. While the high representation could well be the result of high transcript levels in certain libraries derived from particular tissues, many libraries are normalized whereby redundant transcripts from genes with high expression have been eliminated. The presence of a large number of sucrose synthase transcripts in datasets derived from stress cDNA libraries derived from the four abiotic conditions (heat, cold, drought and salt) is not new. High levels of AtSus1 mRNA has been known to be induced by cold (Baud et al., 2004; Dejardin et al., 1999; Martin et al., 1993) and drought (Baud et al., 2004; Dejardin et al., 1999; Pelah et al., 1997). AtSus2 transcripts are reported to increase preferentially due to oxygen deficiency (Dejardin et al., 1999). AtSus5 and AtSus6 are not known to respond to stress (Baud et al., 2004). This is relevant as the current analysis shows that AtSus6 is significantly over-represented in datasets derived from stress.