UBC Faculty Research and Publications

New insights into the evolutionary history of plant sorbitol dehydrogenase Jia, Yong; Wong, Darren C; Sweetman, Crystal; Bruning, John B; Ford, Christopher M Apr 12, 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-12870_2015_Article_478.pdf [ 2.61MB ]
JSON: 52383-1.0074596.json
JSON-LD: 52383-1.0074596-ld.json
RDF/XML (Pretty): 52383-1.0074596-rdf.xml
RDF/JSON: 52383-1.0074596-rdf.json
Turtle: 52383-1.0074596-turtle.txt
N-Triples: 52383-1.0074596-rdf-ntriples.txt
Original Record: 52383-1.0074596-source.json
Full Text

Full Text

RESEARCH ARTICLEnBs tbthe early divergence within the NAD (H)-dependent on sorbitol while also being able to oxidize the other poly-Jia et al. BMC Plant Biology  (2015) 15:101 DOI 10.1186/s12870-015-0478-5atoms are coordinated to the zinc. The water molecule co-ordinating the zinc atom acts a general base and abstractsAustraliaFull list of author information is available at the end of the articlemedium-chain dehydrogenase/reductase (MDR) super-family (with a typical ~350-residue subunit), sharing adistant homology with alcohol dehydrogenase (ADH, EC1.1.1.1) [14-17]. SDH catalyses the reversible oxidation of arange of related sugar alcohols into their correspondingols at lower reaction rates [6,13,18,20]. The process ofsorbitol oxidation by human SDH requires a catalytic zincatom which is coordinated by the side chains of threeamino acids (44C, 69H, 70E, numbering in human SDH)and one water molecular. NAD+ binds to the protein first,followed by sorbitol. The backbone of sorbitol stacksagainst the nicotinamide ring while the C1 and C2 oxygen* Correspondence: christopher.ford@adelaide.edu.au1School of Agriculture, Food and Wine, University of Adelaide, Adelaide 5005,yeasts [5], bacteria [6] and plants [7-13]. It represents(LIDH, EC is involved in tartaric acid (TA) synthesis in Vitis vinifera and is highly homologous to plant SDHs.Despite efforts to understand the biological functions of plant SDH, the evolutionary history of plant SDH genesand their phylogenetic relationship with the V. vinifera LIDH gene have not been characterized.Results: A total of 92 SDH genes were identified from 42 angiosperm species. SDH genes have been highly duplicatedwithin the Rosaceae family while monocot, Brassicaceae and most Asterid species exhibit singleton SDH genes. CoreEudicot SDHs have diverged into two phylogenetic lineages, now classified as SDH Class I and SDH Class II. V. viniferaLIDH was identified as a Class II SDH. Tandem duplication played a dominant role in the expansion of plant SDH familyand Class II SDH genes were positioned in tandem with Class I SDH genes in several plant genomes. Protein modellinganalyses of V. vinifera SDHs revealed 19 putative active site residues, three of which exhibited amino acid substitutionsbetween Class I and Class II SDHs and were influenced by positive natural selection in the SDH Class II lineage. Geneexpression analyses also demonstrated a clear transcriptional divergence between Class I and Class II SDH genes inV. vinifera and Citrus sinensis (orange).Conclusions: Phylogenetic, natural selection and synteny analyses provided strong support for the emergence of SDHClass II by positive natural selection after tandem duplication in the common ancestor of core Eudicot plants. Thesubstitutions of three putative active site residues might be responsible for the unique enzyme activity of V. viniferaLIDH, which belongs to SDH Class II and represents a novel function of SDH in V. vinifera that may be true also of otherClass II SDHs. Gene expression analyses also supported the divergence of SDH Class II at the expression level. This studywill facilitate future research into understanding the biological functions of plant SDHs.Keywords: Sorbitol dehydrogenase, L-idonate-5-dehydrogenase, Gene duplication, Functional divergence, Tartaric acid,Ascorbic acid, GrapevineBackgroundSorbitol dehydrogenase (SDH, EC is commonlyfound in all kinds of life forms, including animals [1-4],ketoses [7,13,18-21], preferring polyols with a d-cis-2,4-dihydroxyl (2S,4R) configuration and a C1 hydroxyl groupnext to the oxidation site at C2, such as sorbitol, xylitoland ribitol (Additional file 1). It exhibits the highest activityNew insights into the evoplant sorbitol dehydrogeYong Jia1, Darren CJ Wong1,2, Crystal Sweetman1,3, JohnAbstractBackground: Sorbitol dehydrogenase (SDH, EC ihigher plants. SDH genes in some Rosaceae species could© 2015 Jia et al.; licensee BioMed Central. ThisAttribution License (http://creativecommons.oreproduction in any medium, provided the orDedication waiver (http://creativecommons.orunless otherwise stated.Open Accesslutionary history ofaseBruning4 and Christopher M Ford1*he key enzyme involved in sorbitol metabolism ine divided into two groups. L-idonate-5-dehydrogenaseis an Open Access article distributed under the terms of the Creative Commonsrg/licenses/by/4.0), which permits unrestricted use, distribution, andiginal work is properly credited. The Creative Commons Public Domaing/publicdomain/zero/1.0/) applies to the data made available in this article,Jia et al. BMC Plant Biology  (2015) 15:101 Page 2 of 23the proton of the C2 hydroxyl, which creates an electronflow to NAD+, leading to the oxidation of sorbitol at C2and the final production of NADH [22].Plant SDH is the key enzyme in the sorbitol metabol-ism pathway [7,13,20,21,23] and has been associatedwith resistance to abiotic stresses such as drought andsalinity. SDH activity regulates the levels of polyols[13,23], which act as important osmolytes duringdrought stress and recovery processes [24]. In Rosaceaespecies sorbitol occurs as the major photosynthate andphloem transported carbohydrate [25]. In these plants,which include apple [26-31], pear [32,33] and loquat[34,35], SDH plays a crucial role in the oxidation ofsorbitol and its translocation to sink tissues such as de-veloping fruits and young leaves. Gene transcript leveland enzyme activity remain high during fruit develop-ment and maturation, dropping gradually in laterstages, and contributing to the sugar accumulation inthe ripening fruits [27-30,34-36]. The role of sinkstrength regulation for SDH is of particular researchinterest given the economic importance of these fruitspecies. Additionally, SDH has been shown to be in-volved in the sugar metabolism process during seedgermination of some herbaceous plants including soy-bean [37] and maize [8,38].Despite efforts to understand the physiological role ofSDH in plants, little attention has been paid toward theevolutionary history of the plant SDH gene family. Thedistribution of the SDH genes in higher plants appearsto be species-dependant. In particular, 9 paralogousSDH genes have been reported in apple [27] and 5 inJapanese pear [39]. In contrast, other plant genomessuch as A. thaliana [23], tomato [11] and strawberry[12] contain only one SDH gene. Recent studies have in-dicated that there are two groups of SDH present insome Rosaceae plants. Park et al. [10] isolated four SDHisoforms (MdSDH1-4) from Fuji apple and found thatMdSDH2-4 could be clearly distinguished fromMdSDH1 based on the deduced amino acid sequence,showing 69–71% identity with MdSDH1 and 90–92%identity with each other. In addition, MdSDH2-4 wereexpressed only in sink tissues such as young leaves,stems, roots and maturing fruits while MdSDH1 washighly expressed in both sink and source organs [10].Nosarzewski et al. [27] identified nine SDHs (SDH1-9)from the Borkh apple genome and showed that all iso-forms except SDH1 (71–73% identity with SDH2-9)were highly homologous with an identity of 91–97%.Similar observations have been made with the SDH iso-forms (PpySDH1-5) identified in pear whereby PpySDH5differed from PpySDH1-4 at both the primary structurelevel and the gene transcriptional level [39]. Preliminaryphylogenetic analyses have classified these homolo-gous SDHs into two groups based on primary proteinstructures [10,29,33,40]. However, these studies focused ononly one or just a few related Rosaceae species. No com-prehensive phylogenetic analysis has been performed onSDH across a broad range of angiosperm species.Gene duplication is widespread in plant genomes.Functional divergence after gene duplication is the majormechanism by which genes with novel function evolve;this phenomenon plays a key role in the evolution ofphenotypic diversity [41-44]. The current understandingof gene evolution via duplication suggests that dupli-cated genes could arise through different mechanismsincluding unequal crossing over (resulting in tandemduplication), retrotransposition, segmental duplicationand chromosomal (or whole genome) duplication[42,45]. Most duplicated genes are lost due to the accu-mulation of mutations that render them non-functional(pseudogenization) [42]. However, they can be retainedunder certain circumstances whereby the acquisitionof beneficial mutations leads to novel function (neo-functionalization), which requires positive natural selec-tion, or through adoption of part of the functions of theancestral gene (sub-functionalization), which could occurby expression divergence or functional specialization ofprotein [41,42,46,47]. The latter usually involves a shift inthe enzyme substrate specificity.Protein structural analyses have shown that the LIDHof V. vinifera, which catalyses the inter-conversion ofL-idonate and 5-keto-D-gluconate (5KGA) in the tar-taric acid (TA) synthesis pathway [48], is highly homolo-gous to plant SDHs, sharing ~77% amino acid sequencesimilarity with SDH from tomato (Gene ID: 778312) andA. thaliana (Gene ID: AT5G51970) [48]. The 366 aminoacid LIDH (UniProt ID: Q1PSI9) contains an N-terminalGroES-like fold and a C-terminal Rossmann fold [48],characteristics of the ADH family [49], which has a distanthomology to SDH [14-17]. However, unlike other plantSDHs, LIDH displays principal activity against L-idonateand has a low reaction rate with sorbitol [48]. The uniquesubstrate specificity of LIDH was suggested to be due tosmall changes in amino acid sequence encoded by paralo-gous genes [48].In this study, a comprehensive phylogenetic analysisof angiosperm SDHs was conducted using currentlyavailable genomic data. A computational approach wasemployed to characterise the natural selection pressureon plant SDH. The protein structures of the SDH homo-logues in V. vinifera were modelled based on humanSDH (PDB:1PL8) to identify the putative active site resi-dues of plant SDHs. Transcription and co-expressiondata of SDH genes were also extracted from recent pub-licly available microarray and co-expression databasesand analysed. New insights into the evolution history ofthe plant SDH family and the evolutionary origin ofV. vinifera LIDH will be discussed.Results and discussionIdentification of sorbitol dehydrogenase (SDH)homologous genes in higher plantsA database homology search identified 92 SDH homolo-gous genes from 42 species (Figure 1; See Additionalfile 2: Table S1 for identified gene IDs and Additionalfile 3 for gene sequences in corresponding species). Atleast one putative SDH gene was present in each plantgenome studied, consistent with previous studies [17]that suggested the ubiquity of SDH and its functionalimportance across all life forms. However, the distribu-tion of SDH homologous genes varied dramaticallyacross species. Monocot species (n = 8) uniformly pre-sented a single SDH gene, and this same observationwas made with Brassicaceae plants (n = 7) from theEudicot group. It was recently reported that there are2 SDH genes in both rice (monocot) and A.thaliana(Brassicaceae) [50], however, in both cases these SDHgenes were found to be alternative transcripts of a singlegene. All except one species from the Asterid clade andthe Leguminosae family had one SDH gene, the excep-tions being Solanum tuberosum (potato) and Glycinemax (soybean), respectively, which both had two copies.By contrast, numerous copies of SDH genes were foundin Rosaceae species, which employ sorbitol as the majortransported carbohydrate [25]. Malus × domestica (apple)contained 16 putative SDH genes, the highest numberamong all species investigated. A previous study [50]identified 17 SDH genes in the apple genome, however,the extra putative SDH (MDP0000506359) was only apartial gene (177 residues) and was excluded from thepresent study. In addition to apple, other Rosaceaelys. SJia et al. BMC Plant Biology  (2015) 15:101 Page 3 of 23Figure 1 Distribution of SDH homologous genes in higher plants. Closemap was based on the total copy number of SDH genes in each specieobtained from literature; additional SDHs may be identified in these two spclassification of SDH Class I and SDH Class II was based on the phylogenetrelated species were specified accordingly. The gene abundance heatDHs of P. bretschneideri [39] and E. japonica (loquat) [35] wereecies when complete genome information becomes available. Theic analysis carried out in the present study.Jia et al. BMC Plant Biology  (2015) 15:101 Page 4 of 23species such as Prunus persica (peach), Prunus mume(Chinese plum), Eriobotrya japonica (loquat) and Pyrusbretschneideri (pear) had 4, 3, 1 and 5 putative SDHgenes respectively. It should be noted that the informa-tion of SDH numbers in loquat [35] and pear [39] was re-trieved from earlier reports, and that more SDH genesmay be found when complete genome data for thesespecies become available. Although Fragaria vesca(strawberry) belongs to the Rosaceae family, only oneSDH gene was present in this species. Unlike other Rosa-ceae fruit species, F. vesca utilizes sucrose instead ofsorbitol as the main translocated carbohydrate [51]. Ac-cording to a recent development in the evolution by du-plication theory, a proper gene dosage should be kept tomaintain a stoichiometric balance in macromolecularcomplexes such as functional proteins, thereby ensuringthe normal functioning of a particular biological process[41,52]. Transportation and assimilation of sorbitol is aRosaceae-specific metabolism. The retention of highlyduplicated SDH genes in Rosaceae species suggests that ahigher dosage of SDH transcription or enzyme activity isneeded to facilitate sorbitol metabolism in these species.Three putative SDH genes were identified in the V. vi-nifera genome. One (GSVIVT01010646001) corre-sponded to the previously characterized LIDH (UniprotNo. Q1PSI9) [48] while the other two shared 99%(GSVIVT01010644001) and 77% (GSVIVT01010642001)amino acid sequence identity with V. vinifera LIDH(Additional file 2: Table S4). Other important crops suchas C. sinensis (orange), Theobroma cacao (cocoa), andPelargonium hortorum (a geranium species) had 3, 2 and2 SDH genes respectively. P. hortorum and S. tuberosumare of particular interest in this study because they havealso been shown to accumulate significant levels of TA,like V. vinifera [53,54]. Another species that should benoted is Aquilegia coerulea (a flower native to the RockyMountains), which belongs to the Eudicot family but hasbeen recognized as an evolutionary intermediate [55] be-tween monocot and core Eudicot plants, and contained7 SDH paralogues.Phylogenetic analysis of plant sorbitol dehydrogenasefamiliesTo determine the evolutionary history of plant SDHfamily and the phylogenetic relationship between LIDHand SDH, a phylogeny of the SDH family was recon-structed. Consistent results were obtained using bothNeighbour Joining (Figure 2A; Additional file 4) andMaximum Likelihood (Figure 2B) methods. As can beseen in the Maximum Likelihood tree (Figure 2B), thetarget proteins divided at the basal nodes into threemajor clusters, corresponding to the three life kingdoms:fungi, animal and plant (Bootstrap supports at 0.98, 1and 1 respectively). The overall topology of the plantSDH clade was in agreement with the Phytozome spe-cies tree (http://www.phytozome.net/), indicating thatthe phylogeny results were reliable. Specifically, monocotplants (n = 8) formed a single clade with strong support(0.91), corresponding to the early split between monocotand dicot lineages. A. coerulea SDHs separated into asingle group (0.91) which positioned itself betweenmonocot and core Eudicot plants. The Aquilegia genusbelongs to the Eudicot order Ranunculales which hasbeen established as a sister clade to the rest of the coreEudicot [56-58] and agrees with the present phylogeneticanalysis.The core Eudicot SDHs split into two distinct lineagesin the Maximum Likelihood tree (Figure 2B). The firstlineage (classified as Class I) covered all core Eudicotspecies included in this study while the second (Class II)had a narrower coverage and was less expanded com-pared to SDH Class I. The divergence of core EudicotSDHs into two lineages was in agreement with previousreports that SDHs from some Rosaceae species couldbe separated into two groups [10,29,33]. All Rosaceaeplants (n = 5) investigated in this study except F. vesca(strawberry) had multiple copies of SDH genes that cov-ered both SDH Class I and SDH Class II. However,within these species, the distribution of SDHs amongthe two SDH classes varied greatly. In particular, 15 outof the 16 SDHs from M. domestica and 4 out of the 5SDHs from P. bretschneideri fell into SDH Class I while3 out of the 4 SDHs from P. persica and 2 out of the 3SDHs from P. mume belonged to SDH Class II. Otherspecies retaining two classes of SDHs included S. tubero-sum,V. vinifera, Eucalyptus grandis, C. sinensis, T. cacao,P. hortorum, Populus trichocarpa, Linum usitatissimum,Jatropha curcas and Manihot esculenta, from differentorders or families. In contrast, Brassicaceae plants(n = 7), Leguminosae plants (n = 4) and Asterid plants(n = 2) except S. tuberosum contained either a singleSDH or two SDHs that could only be classified intoSDH Class I. Within both SDH Class I and Class IIclades, Rosaceae SDHs (except F. vesca) formed separatephylogeny groups (Figure 2B), implying divergent mo-lecular characteristics for SDHs from this family. Mostrecent phylogenetic analyses [59,60] have placed Vitaceaeas a sister clade to the Rosid plants in the core Eudicotgroup. The presence of two classes of SDHs in both V. vi-nifera and S. tuberosum (Asterids) indicated that the di-vergence between SDH Class I and Class II occurredbefore the species radiation of the core Eudicot plants.Moreover, although 7 SDH genes were retained inthe genome of the evolutionarily intermediate speciesA. coerulea, none of them could be classified into SDHClass I or SDH Class II. Taken together, our results sug-gested that SDH Class I and Class II might have divergedduring the common ancestor of core Eudicot plantsFigure 2 (See legend on next page.)Jia et al. BMC Plant Biology  (2015) 15:101 Page 5 of 23but after the branching of the basal Eudicots such asRanunculales. This corresponds to a period of aboutthe Class I branch in the Neighbour Joining tree (Figure 2A;Additional file 4).(See figure on previous page.)Figure 2 Phylogenetic tree showing the evolutionary history of the angiosperm SDH family. A: A simplified schematic phylogeny of the SDHfamily inferred by MEGA 6.0 [97] software using the Neighbour Joining method. Values (as percentage, cutoff value 50) of Internal branch test(1000 replicates) supports are indicated above the corresponding branches. B: The Maximum Likelihood phylogeny of the SDH family developed byMEGA 6.0 [97] software using the selected best-fitting substitution model JTT + G [99]. 1000 times Bootstraping supports (cut off at 0.5) are displayedabove corresponding branch. Closely related species are annotated accordingly. The V. vinifera LIDH (GSVIVT01010646001) is also marked.HJia et al. BMC Plant Biology  (2015) 15:101 Page 6 of 23125Mya ~ 115Mya [55,58].In the Maximum Likelihood tree, the Class II cladewas well-supported and separated from Class I with lon-ger branch length in general (Figure 2B), suggesting ahigher level of amino acid substitution within this clade.In addition, the topology of the Class II clade (exceptthe Rosaceae group) was in good agreement with thespecies tree at Phytozome (http://www.phytozome.net/search.php), with S. tuberosum (Asterids) diverging firstfollowed by V. vinifera and the rest of the rosid species.This indicates that the Class II SDHs have evolved verti-cally within respective species, which lends further sup-port to the suggestion above that SDH Class I and ClassII have existed during the common ancestry of coreEudicot plants. The backbone topology of the more in-clusive Class I clade in the Maximum Likelihood treewas weakly supported (Bootstrap support under 0.5;Figure 2B), in contrast with the strong clusteringsupport for this clade in the Neighbour Joining tree(Figure 2A; Additional file 4). The weak bootstrap sup-port for the topology of SDH Class I may have resultedfrom a lack of amino acid substitution in this clade, asreflected by the short branch length (Figure 2B). Thecalculation of evolutionary distances for plant SDHsrevealed a pair-wise distance under 0.3 in general(Additional file 2: Table S2), sequence alignment showedthat Class I SDHs tend to be more conserved (average se-quence pair-wise identity 83.4%; Table 1) than Class II(79%; Table 1), which means less amino acid substitutionwithin the Class I clade. These results are consistent withthe strong clustering support for the major sub-clades ofTable 1 Amino acid sequence identity between different SDIdentity Class I Class II A. coerulClass I 83.4 (71-99.7) 75.2 (67-83) 78.5 (71-8Class II 79.0 (71-99) 73.2 (68-8A. coerulea 86.7 (83-9MonocotMammalYeastSDH sequences were divided into six groups (Class I, Class II, A. coerulea, Monocot,the present study (Figure 2). The amino acid sequence identity (as percentage) waseach group is presented, followed by the identity range (in bracket).In contrast to the ubiquity of Class I SDHs, the ab-sence of Class II SDHs in some species may be dueto gene loss after duplication, a common mechanismin gene evolution via duplication [42,61]. This alsoindicated that SDH Class II members may not be es-sential for the normal growth of plants, suggesting adivergent function for this class of SDH genes. Inter-estingly, the previously characterized V. vinifera LIDH(GSVIVT01010646001) [48] was grouped into SDHClass II, providing direct support that in at least one caseSDH Class II may have acquired a novel function, in thisinstance its involvement in the synthesis of TA. Whilethe identity of additional functions for Class II SDHs inother species is unknown, support for a role of someClass II SDHs in TA metabolism may be proposed. Onlya few plant families, including Vitaceae, Geraniaceae andLeguminosae have been shown to accumulate significantlevels of TA [54] and the present results showed thatClass II SDHs were present in both Vitaceae and Gera-niaceae. The absence of Class II SDHs in Leguminosaeplants could be explained by the fact that the synthesis ofTA in Leguminosae proceeds via a different pathway,which bypasses the interconversion of L-idonate and5KGA (catalysed by LIDH) [62]. Recent studies have re-vealed that potato [53], citrus fruits [63] and pear [64,65](all containing Class II SDHs) also produce TA, althoughto a lesser degree than V. vinifera. This is consistent withthe potential correlation between Class II SDHs and TAsynthesis. However, it has also been reported that TA isabsent or found only in trace amount in apple [66], andno information is available about the occurrence of TA ingroupsea Monocot Mammal Yeast6) 77.5 (71-83) 48.0 (44-50) 40.9 (38-43)0) 71.0 (67-74) 46.4 (43-49) 39.3 (37-42)9.7) 75.7 (72-79) 48.0 (47-50) 41.4 (40-43)88.4 (86-93) 47.4 (46-49) 41.5 (40-45)87.8 (82-99.8) 42.3 (39-44)65.5 (48-99.7)Mammal and Yeast SDHs) according to the phylogenetic analysis carried out inobtained using all-vs-all BLAST tool. The average pair-wise identity betweenJia et al. BMC Plant Biology  (2015) 15:101 Page 7 of 23peach even though three copies of Class II SDH geneswere identified in this species (Figure 1). It is possiblethat Class II SDHs have evolved varied functions to meetthe different environmental challenges faced by respect-ive plants. In this context, it would also be valuable forfuture work to investigate the in-planta function of SDHand the occurrence of TA in the evolutionarily intermedi-ate plant A.coerulea, for which 7 SDH paralogues wereidentified.Sequence alignment and protein subdomain analysisSequence alignment and protein subdomain analyseswere performed to investigate the molecular characteris-tics of plant SDHs. Results showed that plant SDHsshared an overall identity above 67% (Table 1), whilehaving ca 48% and ca 41% identities with mammal andyeast SDHs respectively (Additional file 2: Table S4).Plant SDHs were clustered into four groups in thepresent phylogenetic analysis: monocot SDH, A. coeruleaSDH, core Eudicot SDH Class I and SDH Class II. Pro-tein BLAST results showed that Class I and Class IISDHs within the same species generally had an inter-class identity of around 70% and an intra-class identityabove 90% (Additional file 2: Table S4). When comparedwith monocot and A. coerulea SDHs, Class I SDHs al-ways demonstrated a significantly higher similarity thanClass II SDHs (77.5% vs 71.0% and 78.5% vs 73.2% re-spectively; Table 1), suggesting that core Eudicot Class ISDHs have a closer distance to monocot and A. coeruleaSDHs and that SDH Class II may have diverged fromSDH Class I. In addition, Class I SDHs tend to be morehomologous than Class II SDHs (83.4% vs 79.0%;Table 1). No significant difference between the two SDHclasses was observed when compared to mammal oryeast SDHs (48.0% vs 46.4% and 40.9% vs 39.3% respect-ively; Table 1). Protein functional domain predictionidentified two functional domains for plant SDHs: anN-terminal GroES-like fold and a C-terminal Rossmannfold (Figure 3; See Additional file 5 for the complete se-quence alignment). Secondary structure analysis showedthat these two domains tended to be highly conservedamong all plant SDHs, and amino acid substitutionsmainly occurred at boundary regions linking secondarystructural elements such as alpha-helices and beta-sheets(Figure 3).Gene duplication pattern characterization and syntenyanalysisTo characterise the expansion patterns of plant SDHgene family, nine species that were from different fam-ilies and contained both classes of SDHs were selectedfor gene duplication and synteny analyses (C. sinensis,E. grandis, P. mume, P. persica, Populus trichocarpa,M. domestica, S. tuberosum, T. cacao and V. vinifera). Asshown in Table 2 (See Additional file 6 for the originaloutput data), tandem duplication contributed themost to the expansion of the core Eudicot SDH fam-ily, followed by WGD/Segmental duplication. Dis-persed SDHs (MDP0000305455, MDP0000759646 andPGSC0003DMC400055323) and a single proximal SDH(MDP0000188054) were identified only in M. domesticaand S. tuberosum. Based on phylogenetic classification inthe present study, Class I and Class II SDH genes fromE. grandis, P. trichocarpa, T. cacao and V. vinifera arelocated in a tandem manner in their correspondingchromosomes, which provides strong support that SDHClass I and SDH Class II are tandem duplications. Asimilar pattern was observed with C. sinensis wherebyCs9g16660.1 (SDH Class II) is separated by a single-gene insertion with the two Class I SDH genes(Cs9g16680.1, Cs9g16690.1; data not shown). This maybe caused by gene insertion after tandem duplication.Class I and Class II SDH genes in the three Rosaceaespecies (M. domestica, P. mume, P. persica) and inS. tuberosum are separated either on the one chromo-some or on separate chromosomes altogether, indicat-ing a divergent evolutionary history for SDH genes inthe Rosaceae family and in S. tuberosum compared toother plants. SDH genes on chromosome 1 (md1) andchromosome 7 (md7) in M. domestica were highly du-plicated by tandem duplication (Table 2), in contrastto the other Rosaceae species (P. mume, P. persica).Notably, the Class I SDH gene from S. tuberosum(PGSC0003DMC400055323) and the Class II SDHgene from M. domestica (MDP0000305455) were iden-tified as dispersed duplicates, which may underpin thedivergent sorbitol metabolism profiles across thesespecies.To investigate the conservation of SDH genes acrossspecies, collinear SDH gene pairs were identified withinand across species. SDH genes from the nine above-mentioned species were analysed. The single SDH gene(AT5G51970) from the model plant A. thaliana was alsoused as a reference for collinear block identification. Asshown in Figure 4, all target plant genomes contained atleast one SDH gene (corresponding to chromosome po-sitions A, B, C, D, E, H, J, L, N, P and Q in Figure 4)with collinear SDH genes in all other nine species stud-ied, indicating a conserved collinear SDH block. SDHgenes at gene positions F, G, I, K and O, concerning onlythe Rosaceae species investigated, were collinear withSDH genes in only some of the species included in thepresent analysis. In particular, position F at chromosome8 (pp8) of P. persica paired only with position I atchromosome 6 (Pm6) of P. mume. While position F wasfound collinear only with position I, position I had an-other collinear region at position O from E. grandis.Position G at chromosome 4 (pp4) of P. persica wasFigure 3 Multiple sequence alignment of plant SDH family. ESPript output was obtained with the sequence alignment of plant SDHs and humanSDH. Secondary structures were inferred using human SDH (PDB: 1PL8) as a template, with springs representing helices and arrows representingbeta-strands. Sequences were grouped into 1 (1PL8 and core Eudicot SDH Class I), 2 (core Eudicot SDH Class II), 3 (A.coerlea SDH) and 4 (monocotSDH). Amino acid site numbering above the alignment is according to LIDH (Q1PSI9) without the first 20 amino acids. Adjacent similarity aminoacid sites were boxed in blue frame. Similarity calculations were based on the complete SDH alignments but only partial sequences for SDH Class Iand SDH Class II were displayed. The active site residues identified in this study are marked with red triangles. Conserved domains are indicatedabove the alignment.Jia et al. BMC Plant Biology  (2015) 15:101 Page 8 of 23Jia et al. BMC Plant Biology  (2015) 15:101 Page 9 of 23Table 2 Gene duplication patterns of plant SDHSpecies Chromosome ID SDH gene IDC. sinensis cs9 Cs9g16680.1 (orange1.1g017426m)only paired with positions A, E and K from A. thali-ana, P. trichocarpa and M. domestica respectively.Some collinear SDH gene pairs, such as F-I, G-K andK-O, were restricted to Rosaceae species only, reflect-ing genetic features shared only by these plants. Not-ably, intra-species collinear SDH pairs were identifiedcs9 Cs9g16690.1 (orange1.1g048013m)cs9 Cs9g16660.1 (orange1.1g017793m)E. grandis eg11 Eucgr. K00213.1eg11 Eucgr.K00212.1M. domestica md1 MDP0000786110md1 MDP0000873573md1 MDP0000707567md1 MDP0000515106md1 MDP0000250546md1 MDP0000874667md1 MDP0000638442md1 MDP0000123910md1 MDP0000305455md7 MDP0000188052md7 MDP0000171573md7 MDP0000188054md7 MDP0000167088md7 MDP0000807470md14 MDP0000759646P. mume Pm5 Pm019393Pm6 Pm021180Pm6 Pm021179P. persica pp2 ppa007458m|PACid:17644502pp4 ppa007327m|PACid:17655491pp8 ppa007343m|PACid:17644328pp8 ppa007374m|PACid:17655656P .trichocarpa pt12 POPTR_0012s13780pt12 POPTR_0012s13790S. tuberosum st01 PGSC0003DMC400055323st06 PGSC0003DMC400043871T. cacao tc03 Tc03_g019280tc03 Tc03_g019270V. vinifera vv16 GSVIVT01010642001vv16 GSVIVT01010646001vv16 GSVIVT01010644001SDH gene duplication patterns were characterized by the duplicate_gene_classifier pDuplication or segmental duplication. “SDH Class” is defined according to the presenot be anchored in any chromosome and was therefore absent in this table.SDH class Duplication pattern Start position End positionI Tandem 16143063 16147624only within M. domestica but not in P. mume, P. per-sica and S. tuberosum although all of these specieshave SDH genes located on multiple chromosomes(Figure 4; See Additional file 2: Table S5 for identifiedcollinear SDH gene pairs). This observation could beexplained by the fact that the apple genomeI Tandem 16150122 16154404II WGD or Sgm 16135216 16138066I Tandem 2624187 2627945II Tandem 2615486 2618589I Tandem 25191824 25193641I Tandem 25182502 25183812I Tandem 25180931 25182241I Tandem 25177288 25178612I Tandem 25173127 25174375I Tandem 25157544 25158783I WGD or Sgm 25149134 25150444I WGD or Sgm 25087036 25088743II Dispersed 14150327 14159200I Tandem 23301490 23302735I WGD or Sgm 23281847 23283529I Proximal 23310942 23312187I Tandem 23405354 23406795I WGD or Sgm 23390960 23392683I Dispersed 24043122 24044360I WGD or Sgm 23673441 23675177II Tandem 7217228 7219256II Tandem 7217228 7225304I WGD or Sgm 24766424 24768515II WGD or Sgm 17729024 17731238II Tandem 15254677 15256888II Tandem 15249947 15251989II WGD or Sgm 13789342 13787442I WGD or Sgm 13790093 13792804I Dispersed 1594220 1598967II WGD or Sgm 24156879 24158593I WGD or Sgm 18300080 18303115II WGD or Sgm 18298897 18296706I WGD or Sgm 15653874 15651701II Tandem 15675560 15678887II Tandem 15666264 15664425rogram in the MCScanX package. “WGD or Sgm” refers to Whole Genoment phylogenetic analysis. Notably, MDP0000149907 from M. domestica couldJia et al. BMC Plant Biology  (2015) 15:101 Page 10 of 23underwent a recent (>50Mya) WGD, which doubledthe chromosome number from nine to 17 in thePyreae [50] while most other Rosaceae plants have ahaploid chromosome number of 7, 8 or 9. S. tubero-sum was unique among the species investigated in thatit had a Class II SDH gene (PGSC0003DMC400043871)but no Class I SDH gene preserved in the collinearregion (Figure 4). The Class I SDH gene(PGSC0003DMC400055323), which was identified asa dispersed duplication (Table 2), was the only SDHgene for which no collinear gene was identified in thepresent analysis. Since the Class II SDH homologueFigure 4 Identification of collinear gene pairs among plant SDH families. Alinked by red curved lines. SDH genes located at each position in correspogenomic collinearity background. Only those chromosomes containing SD(LIDH) in V. vinifera has been shown to be involved inTA synthesis [48], it would be of great interest to investi-gate the potential role of SDHs in S. tuberosum, whichhas also been shown to accumulate a significant amountof TA [53]. Noteworthy, S. lycopersicum, another speciesfrom the Solanale order, accumulates no TA [67] andcontains only a single SDH, which belongs to Class I(Figure 2B).Natural selection analysisAssessment of synonymous and non-synonymous substi-tution ratios is important to understand molecularcircular plot of SDH gene family collinearity. Collinear SDH genes arending chromosomes are indicated. Family collinearity is shown in theH genes are included.evolution at the amino acid level [68,69]. To examinethe intensity of natural selection acting on the specificclade, the ratio (w) of non-synonymous substitution tosynonymous substitution in the developed plant SDHphylogeny was investigated, whereby w<1, w=1 andw>1 indicated purifying selection, neutral evolutionand positive selection respectively. Based on ourphylogeny results, four branches (“monocot SDH”,“A. coerulea SDH”, “core Eudicot SDH Class I” and“core Eudicot SDH Class II”) were specified for w as-sessments (w [mono], w [Aer], w [sdhC1] and w[sdhC2] respectively). Firstly, the branch-specific like-lihood model [70] was applied to the SDH data. Ascan be seen in Table 3, Likelihood-ratio tests (LRT)showed that the two-ratio model and the four-ratiomodel fit the dataset significantly better (2Δl = 12.6with p = 0.0004, df = 1 and 2Δl = 13.2 with p = 0.0042,df = 3 respectively) than the one-ratio model. In con-trast, the three-ratio model assumption lacked statis-tical support (2Δl = 0.2 with p = 0.9048, df = 2). Giventhat the two-ratio and four-ratio models assume un-equal w ratios for the Class I and Class II brancheswhile the three-ratio model specifies w(sdhC1)=w(sdhC2) (Table 3), the above calculation suggestedthat the w ratio for the core Eudicot SDH Class II wasfour-ratio model, which assumes unequal w ratios for themonocot, A.coerulea and Class I branches (Table 3), wasnot significantly better (2Δl = 0.6 with p = 0.7408, df = 2)than the two-ratio model (assuming uniform ratio forthese branches; Table 3). This indicated that the w ratiosfor monocot, A. coerulea and core Eudicot Class Ibranches had no significant difference. Notably, allbranch-specific models tested demonstrated a low wvalue for the monocot, A. coerulea and Class I branches(w[mono]=w[Aer]=w[sdhC1]=0.10415 with the two-ratiomodel and w[mono]=0.10428, w[Aer]=0.09731, w[sdhC1]=0.0001with the four-ratio model), suggestingthat plant SDHs have been under strong purifying se-lection. This agrees well with the suggestion thatfunctional proteins are usually under strong structuraland functional constraints [71]. It should be notedthat w[sdhC2] were infinite in both multi-ratio models(w[sdhC2]=859 and 999 respectively). This is becausean extremely low level of synonymous substitution orno synonymous substitution was detected in the SDHClass II clade. On the other hand, the number of non-synonymous substitutions in the core SDH Class IIclade was estimated to be 12.7 and 12.8 respectively forthe two-ratio model and the four-ratio model. In con-trast, only 0.4 non-synonymous substitution was detectedf(A(A59.10(s.1099(A(p, p(w, p1, p113ereamls,Jia et al. BMC Plant Biology  (2015) 15:101 Page 11 of 23significantly different from that of Class I. Moreover, theTable 3 Natural selection tests of plant SDHModel np l = ln L Estimates oM0: one-ratiow(mono)=w (Aer)=w(sdhC1)=w(sdhC2) 1 -30147.4 w(mono)=wBranch-specific modelsw(mono)=w(Aer)=w(sdhC1)≠w(sdhC2)(two ratios)2 -30141.1 w(mono)=ww(sdhC2)=8w(mono)≠w(Aer)≠w(sdhC1)=w(sdhC2)(three ratios)3 -30147.3 w(mono)=0w(sdhC1)=ww(mono)≠w(Aer)≠w(sdhC1)≠w(sdhC2)(four ratios)4 -30140.8 w(mono)=0w(sdhC2)=9w(mono)=w(Aer)=w(sdhC1)≠w(sdhC2)(two ratios with w(sdhC2) fixed to 1)1 -30141.4 w(mono)=wSite-specific modelsM1:Neutral (2 site classes) 2 -29650.0 p0=0.87775M2:Selection (3 site classes) 3 -29650.0 p0=0.87775w0=0.07628Branch-site models (SDH Class II as foreground lineage)Model A Null (4 site classes) 3 -29643.2 p0=0.33951Model A (4 site classes) 4 -29640.9 p0=0.82864(w1=1), w2=All calculations were implemented using codeml at PAML4.7. Different models wparameters, “l = (ln L)” refers to the log value of the likelihood. The estimated parsite classes respectively. In the one-ratio model M0 and the Branch-specific modeA. coerulea, SDH Class I and SDH Class II branches respectively. In the Site-specific mospecific site classes in respective models (see the Methods section for more details). Fobranch. Amino acid site numbering is according to LIDH (Uniprot No: Q1PSI9) withoutfor the SDH Class I clade with the two-ratio modelparameters Positively selected siteser)=w(sdhC1)=w(sdhC2)=0.10492 Not Allowed (NA)er)=w(sdhC1)=0.10415,.33956NA510, w(Aer)=0.10821,dhC2)=0.06935NA428, w(Aer)=0.09731, w(sdhC1)=0.0001, NAer)=w(sdhC1)=0.10424 (w(sdhC2)=1) NA1=1-p0=0.12225); w0=0.07628 (w1=1) NA1=0.07499 (p2=1-p0-p1=0.04726);1=1), w2=1None=0.04783 (p2+p3=0.61266); w0=0.07544 NA=0.11666 (p2+p3=0.0547), w0=0.075442.6226Sites for foreground lineage:42H,43F,112G, 113S,116T, 270Q(p > 0.99);specified according to the software instruction. “np” refers to the number ofeters w and p refer to the Ka/Ks ratio and the percentage of the correspondingw(mono), w(Aer), w(sdhC1) and w(sdhC2) stand for the w ratios for the monocot,dels and the Branch-site models, w0, w1 and w2 represent the w ratios for ther the Branch-site models, the SDH Class II branch was specified as the foregroundthe first 20 amino acids.Jia et al. BMC Plant Biology  (2015) 15:101 Page 12 of 23(Additional file 7: branch-specific-two-ratio-output) andno non-synonymous substitution was detected with thefour-ratio model (Additional file 7: branch-specific-four-ratio-output). These results provided clear evidence thatpositive selection had occurred in the lineage leading tocore Eudicot SDH Class II. To test whether w[sdhC2] issignificantly higher than 1, the log likelihood value(Table 3; Additional file 7: branch-specific-two-ratio-null-output) was calculated for the two-ratio model withw[sdhC2]=1 fixed. Results showed that this model wasnot significantly worse than the two-ratio model withoutthe “w[sdhC2]=1” constraint (2Δl = 0.6 with p = 0.4386,df = 1), suggesting that w[sdhC2] was not significantlygreater than 1 at the 5% significance level. This leads tothe hypothesis that positive selection in SDH Class IImight have only affected particular amino acid residuesin the protein sequence, which is possible for a functionalprotein under strong structural and functional con-straints [72]. To test this, Site-specific likelihood analysiswas performed on the same data, which assumes variableselection pressures among amino acid sites but no vari-ation among branches in the phylogeny. Results (Table 3:model M2) showed that the selection model (M2) fit-ted the dataset significantly better (2Δl = 994.8 withp = 0.0001, df = 2) than the one-ratio model but wasnot better (2Δl = 0 with p = 1, df = 1) than the neutralmodel (M1). These results indicated a significant vari-ation of selection pressure among amino acid sites ofplant SDH. However, the Selection model failed to detectany positively selected amino acid site at a significantlevel (Table 3; Additional file 7: site-specific-output),which suggested that no positively selected amino acidsite could be identified across all branches. Therefore, wespeculate that the positive selection might have onlyacted on a few amino acid sites in the core Eudicot SDHClass II clade.In this context, a Branch-site model [73] that permitsvariable w ratios among both amino acid sites andbranches was applied. Model A successfully identifiedthe potential amino acid sites under positive selection inthe SDH Class II branch (Table 3; Additional file 7:branch-site-modelA-output). Specifically, 42H, 43F, 112G,113S, 116T and 270Q (numbering in LIDH (Q1PSI9)without the first 20 amino acids) were identified withModel A (Bayes Empirical Bayes analysis possibility >0.99;Additional file 7: branch-site-modelA-output). LRTs testshowed that Model A fit the data significantly better(2Δl = 18.2 with p = 0.0001, df = 2) than the neutralmodel M1. The comparison (2Δl = 4.6 with p = 0.0320,df = 1) of Model A with its null hypothesis which as-sumes w2=1 (Additional file 7: branch-site-modelA-null-output) indicated that these amino acid sites hadundergone positive selection in SDH Class II but notin the background branches. In addition, the Model A testdemonstrated that 82.90% (model A: p0 = 0.82864; Table 3)of the amino acids of SDH were under strong purifying se-lection (model A: w0=0.07544; Table 3) and 11.7% wereunder neutral selection (model A: p1=0.11666, w1=1;Table 3) in all branches. No positive selection could be de-tected in the background branches (Additional file 7:branch-site-modelA-output). Taken together, these calcula-tions demonstrated that plant SDHs were under strongpurifying selection pressure and were highly conservedacross all the plant species, and more importantly, thatpositive natural selection had occurred in the SDH Class IIclade, affecting specific amino acids, namely 42H, 43F,112G, 113S, 116T and 270Q.Ancestral sequence reconstruction and evolutionrate analysisTo characterize the evolutionary rates for different groupsof plant SDHs, ancestral amino acid sequences for thedeveloped SDH phylogeny were reconstructed. Results(Additional file 8: ancestral-sequence-construction-out-put) showed that 9 potential amino acid substitutions(Y42H, L43F, A112G, T113S, V116T, Q228K, H270Q,N271S, R283A; numbering in LIDH (Q1PSI9) without thefirst 20 amino acids) occurred in the branch leading toSDH Class II from the common ancestor of core EudicotSDH. This finding corresponded well with the natural se-lection analysis, whereby six out of the nine amino acidsites were identified to be under positive selection (42H,43F, 112G, 113S, 116T and 270Q; Table 3). In contrast,no substitution was detected in the branch leading tocore Eudicot SDH Class I (Additional file 8: ancestral-sequence-construction-output and interpreted-ancestral-sequences.fasta). Relative rate tests (RRT) [74] usingmonocot SDH as the out-group showed that core EudicotSDH Class II evolved significantly faster than core EudicotSDH Class I (Additional file 9: ClassI-vs-ClassII.txt), indi-cating a relaxed selection pressure on SDH Class II. Incontrast, A. coerulea SDH and core Eudicot Class I SDHdemonstrated no significant difference (Additional file 9:Aer-vs-ClassI.txt).Protein structure modelling analysisTo deduce the reaction mechanism and identify the po-tential active sites of plant SDHs, protein structuremodels of V. vinifera Class I SDH (Vv_SDH, UniProtNo: D7TMY3) and Class II SDH (Vv_LIDH, UniProtNo: Q1PSI9) were created based on human SDH (PDB:1PL8; 46 ~ 47% identity with Vv_SDH and Vv_LIDH). Li-gands including zinc, NAD+, D-sorbitol and L-idonatewere docked into the models (Additional file 10). Ourmodels contain one zinc binding site, located in the ac-tive site. Some published SDH crystal structures (eg.PDB: 1E3J) contain a second, structural zinc-binding sitedistant from the active site catalytic zinc atom; this isnot however a universal feature of these enzymes. Nofunction has been correlated with the second, structuralzinc-binding site. The sequence of our homology modelsdoes not support a second, structural zinc-binding site,as the necessary side chains required for zinc coordin-ation are absent. A ribbons diagram of the overall struc-ture of the homology models can be seen in Figure 5A,with Vv_SDH and Vv_LIDH adopting a typical dehydro-genase fold with an NAD+ binding site conforming to aRossmann fold. The catalytic zinc ion in the active sitewas modelled coordinating to 36C, 61H and 62E(Figure 5C; numbering in LIDH (Q1PSI9) without thefirst 20 amino acids). All three of these residues togetherwith 147E (corresponding to 155E in human SDH,mediating the water molecule coordinating the zinc atom[22]) are strictly conserved in plant SDHs (Figure 3). The2′ and 3′ hydroxyls of the NAD+ ribose in our modelwere poised to 195D (203D in human SDH), potentiallyforming hydrogen bonds (Additional file 10: Asp195-NAD.png). The preservation of 195D instead of 195A atthis amino acid site has been shown to be the structuralbasis for the selection of NAD (H) over NADP (H) as co-enzyme [75]. This amino acid site is strictly conserved inall plant SDHs (Figure 3), implying that plant SDHspreferably utilize NAD (H). This suggestion is consistentwith the lack of NADP-SDH activity for plant SDHs[7,10,11,13]. Previous characterizations of SDHs fromArabidopsis [13], tomato [11], apple [7,76] and pear [20]A BC DNAD+ NADHL-idonate5KGActioreas oinJia et al. BMC Plant Biology  (2015) 15:101 Page 13 of 23Figure 5 Homology models of Vv_LIDH and Vv_SDH and proposed rea(green) and Vv_SDH_sorbitol (yellow) in Ribbon forms. B. The proposed5-keto-D-gluconate (5KGA). C. Superimposition of the active site residuecorresponding atoms are labelled. Target active site residues are shownat Y42H between Vv_LIDH (green) and Vv_SDH (yellow) with red and whithydrophobicity respectively. (All amino acid site numbering is according ton mechanisms. A. Structure superimposition of Vv_LIDH_idonatection mechanism for Vv_LIDH on the oxidation of L-idonate intof Vv_LIDH (green) and Vv_SDH (yellow). The distances (Å) betweenstick forms and labelled correspondingly. D. Hydrophobicity variancee colours representing the highest hydrophobicity and the lowestLIDH (UniProt No: Q1PSI9) without the first 20 amino acids).Jia et al. BMC Plant Biology  (2015) 15:101 Page 14 of 23have suggested that plant SDHs exhibit highest activityfor the oxidation of sorbitol, while also being able tooxidize other polyols such as xylitol and ribitol at lowerreaction rates. However, the characterization of V. vinif-era LIDH showed that this enzyme demonstrated thehighest reaction rate on L-idonate but had a low reactionrate with sorbitol [48]. Upon docking of L-idonate, wefound overall similar hydrogen bonding patterns withsorbitol as those proposed by Pauly et al. [22] andYennawar et al. [77]. Earlier studies on enzyme sub-strate specificity also indicated that SDHs preferen-tially use substrates with a d-cis-2,4-dihydroxyl (2S,4R)configuration [6,13,18,20] (Additional file 1). L-idonateand D-sorbitol have the same molecular configurationfrom C1 to C4 and differ only at C5 (D and L chirality)and C6 (a hydroxyl group in sorbitol is replaced by a carb-oxyl group in L-idonic acid) (Additional file 1). Proteinmodelling analyses showed that L-idonate occupied a com-parable position in the active site to sorbitol (Figure 5C).Therefore a similar reaction mechanism for L-idonateoxidation by V. vinifera LIDH is possible with D-sorbitoloxidation by human SDH [22]. The hydroxyl groups at C1and C2 of L-idonate were modelled within interactingdistance of the zinc atom in V. vinifera LIDH (Additionalfile 10: C1-C2-Zn.png), which may facilitate the protontransfer from C2 hydroxyl to NAD+, ultimately resulting inan oxidized C2 with ketone and the production of NADH(Figure 5B). Previous work suggested that the preferentialbinding of L-idonate over sorbitol seen in V.vinifera LIDHmay be attributed to amino acid substitution at the cata-lytic sites between paralogous proteins [48]. As a result, thecatalytic site of plant SDHs was investigated based on ourmodels of V.vinifera SDH homologs.Nineteen putative active site residues (36C, 38S, 39D,42H, 48C, 49A, 51F, 61H, 62E, 110F, 112G, 113S, 147E,148P, 151V, 268L, 291F, 292R and 293Y; numbering inLIDH(Q1PSI9) without the first 20 amino acids) wereidentified either coordinating the zinc ion or formingpotential non-covalent interactions with NAD(H) andL-idonate. Ten out of the 19 residues were consid-ered strictly conserved throughout all plant SDH forms,and six additional residues are also largely conserved withvariations in only a few SDH sequences (Figure 3). Theseobservations revealed a potential structural basis for thepreserved function of plant SDHs. Interestingly, threeother residues were found to be uniformly exchanged(Y42H, A112G and T113S) between core Eudicot SDHClass I and Class II while monocot and A. coerulea SDHsresemble SDH Class I at these amino acid sites (Figure 3).A closer inspection of these residues showed that the oxy-gen atom of C5 hydroxyl of L-idonate was poised to po-tentially interact with both 42H and 113S within distancesof 4 Å and 2.6 Å respectively (Figure 5C). Additionally,the oxygen atom of the C6 ketone group of L-idonate waswithin non-covalent interaction distance to 113S (3.5 Å;Figure 5C). Notably, the replacement of 42Y (hydrophobicaromatic side chain) with 42H (charged side chain) inLIDH has the potential to change the hydrophobicity inthe substrate-binding pocket (Figure 5D), which may leadto the preferential binding of L-idonate over D-sorbitol.These observations potentially provided a structural ex-planation for the unique activity of V. vinifera LIDH com-pared to other plant SDHs. Previous studies haveindicated that the chiral configuration at C5 is not a deter-mining factor for SDH substrate specificity [18,20], how-ever, our analysis suggested that the C5 hydroxyl groupand the C6 ketone group of L-idonate potentially affectsubstrate binding affinity due to amino acid substitutionsat 42H, 112G and 113S in Class II SDHs. A previouslyidentified SDH from apple fruit [9] was found to be thesingle Class II SDH (MDP0000305455) in M. domestica inthe present study. This SDH has a much lower affinity forsorbitol (Km 247 mM [9]) compared to other SDHs puri-fied (Km 40.3 mM [76], 86.0 mM [7]) or cloned (Km83.0 mM [10]; SDH Class I) from apple species. While thekinetic differences were suggested to be due to proteinconfiguration changes between the fusion protein and na-tive protein [9], the present analysis indicated that theymight have been be due also to amino acid substitutionsat the catalytic site.From an evolutionary point of view, amino acidchanges leading to the shift of enzyme substrate specifi-city are usually derived from positive Darwinian selec-tion after gene duplication [41,43]. Results from thenatural selection analyses in the present study are con-sistent with this suggestion. The three amino acid sites(42H, 112G and 113S) displaying substitutions betweenSDH Class I and Class II are all under positive naturalselection (Table 3). At the moment, the enzymaticcharacterization of plant SDH is still fragmentary; no in-formation is available regarding plant SDH activity withL-idonate, except for the activity of V. vinifera LIDH[48]. Site mutation and enzymatic studies are cur-rently underway in our laboratory to investigate thishypothesis.Meta-analysis of sorbitol dehydrogenase related geneexpressionIn addition to changes in enzyme activity, gene evolutionafter duplication can also occur at the transcriptionallevel [42]. Expression division appears to be more com-mon than structural evolution and often occurs rapidlyafter gene duplication [42,78,79]. To further characterizethe evolutionary pattern of plant SDH genes and also toexplore the role of SDH related genes during plant de-velopment, a survey of transcriptional data was under-taken. Based on the availability of microarray and RNAsequencing data and the presence of both classes ofJia et al. BMC Plant Biology  (2015) 15:101 Page 15 of 23SDH in the genome, grapevine and citrus species wereselected. In addition, the expression profile of the singleClass I SDH (AT5G51970, Figure 2) in A. thaliana wasused as a model reference [80]. This gene was highlyexpressed in cotyledons, leaves and late stages of seeddevelopment compared to organs such as flowers(stamen, petal, carpel) and shoots (inflorescence, vegeta-tive, transition), where it was marginally expressed (datanot shown).The results support a potential role for SDHClass I during seed germination in A. thaliana [23], soy-bean [37] and maize [8,38]. In grapevines, transcriptionalpatterns of VIT_16s0100g00290 (SDH Class II, LIDH)and VIT_16s0100g00300 (SDH Class I, SDH) were ana-lysed using the normalised grapevine gene expressionatlas of the ‘Corvina’ cultivar [81]. Notable differences ingene expression intensities and dynamics were observedbetween SDH Class I and Class II (Figure 6A; Additionalfile 11: Table S1). The transcript abundance of grapevineSDH Class I was highest in the ripening stages of berries(measured in pericarp, pulp, seeds and skins), resemblingthe expression profiles reported for Class I SDHs inapple [10,27,29]. In most cases, transcript abundancewas lowest in young berry growth stages and increasedgradually until harvest in berry tissues. Developmentalup-regulation of SDH Class I transcripts in other culti-vars such as ‘Shiraz’ [82] and ‘Tempranillo’ [83] duringberry development under normal conditions was alsoevident. In addition, the latter work showed sorbitol ispresent in leaves and berries, and that the biochemicalactivity of SDH Class I, involving sorbitol oxidation, co-incided with SDH class I transcripts levels in these ber-ries during development [83]. Similarly, developmentalincreases of the grapevine SDH Class I transcript wereobserved in leaf, rachis, seed and tendrils. Interestingly,gene expression of grapevine SDH Class I was highlyinduced in winter buds and followed a gradual down-regulation during dormancy release. A similar geneexpression and protein activity pattern reported in rasp-berry [84] and pear [39] respectively may reflect a re-sponse to the environment where dormancy periodsencompasses dehydration and temperature (cold) stress,although developmental processes could take place con-currently. Taken together, this suggests an active role forSDH Class I in developmental processes through the co-ordinated regulation of transcript and protein activitiesin controlling the flux of sorbitol (and related polyols) ingrapevines which may be critical in maintaining cell andtissue homeostasis in the mature tissues [83] where oxi-dative stress is inherent [85,86].Expression profiles of SDH Class II were well repre-sented in most grapevine organs with the highest expres-sion in berries at fruit-set and in flower carpels. Astriking developmental down-regulation of grapevineSDH Class II genes was evident in most grapevineorgans, where expression levels in young tissues of ber-ries (pericarp, flesh, skin and seed), buds, leaves, stemsand tendrils were high and gradually decreased duringdevelopment (Figure 6A). We have previously demon-strated in a cross-comparison study involving RNA-seq,microarray and qRT-PCR in young, early veraison, lateveraison and ripening berries of grapevine [82] that SDHClass II genes were developmentally down-regulatedconsistently in all profiling platforms. This distinct ex-pression coincides with the accumulation of TA biosyn-thesis in young/immature tissues [48,87].In citrus, SDH Class I and SDH Class II geneswere represented by probesets “Cit.9778.1.S1_s_at”and “Cit.9780.1.S1_s_at” respectively. Although geneexpression studies encompassing developmental series incitrus are not as comprehensive compared to A. thalianaand grapevine, several striking observations could be in-ferred (Figure 6B; Additional file 11: Table S1). The citrusSDH Class I gene was highly expressed regardless of organand tissue, including stems, roots, leaves, ovules and fruittissues (albedo, flavedo, juice sacs), similar to that ofgrapevine SDH Class I. Interestingly, SDH Class II geneswere expressed to a very low level (possibly in fact notat all) in the majority of organs, including fruit tissues,except for the root where expression was highest. It isspeculated that this may reflect the trace amount of TAdetected in fruits of sweet oranges and other citrus spe-cies [63]. Until now, no information, to our knowledge,has been reported on the function of citrus SDHs.Given the novel transcription profiles of one the twocitrus Class II SDHs (specifically expressed in roottissues), and the presence of an additional Class II SDH(albeit this sequence was not represented in the arrayfrom which these data were analysed), these featuresmay indicate a novel function of SDHs specific to roottissues of sweet oranges and therefore, deserve more at-tention in future research. In addition to V. viniferaand citrus, divergent transcription profiles have alsobeen reported for SDHs from apple [10] and pear [39]where the single copy Class II SDH genes were shownto be under independent transcriptional regulation fromother SDH genes. Taken together, divergent expressionprofiles for SDH Class I and SDH Class II appear to betrue to all species where two classes are present, sup-porting a gene functional divergence at the expressionlevel.Gene co-expression mining in various plant speciesGene co-expression network analysis (GCA) is based onthe principle that genes involved in similar and/or re-lated biological processes may be expressed in a propor-tional manner, thereby providing a unique tool tounderstand gene function. Based on information avail-ability, co-expressed gene lists of SDHs from A. thaliana,Class IClass IIBerryPericarp-FSBerryPericarp-PFSBerryPericarp-VBerryPericarp-MRBerryPericarp-RBerryFlesh-PFSBerryFlesh-VBerryFlesh-MRBerryFlesh-RBerrySkin-PFSBerrySkin-VBerrySkin-MRBerrySkin-RSeed-FSSeed-PFSSeed-VSeed-MRClass IClass IIBud-LBud-WBud-SBud-BBud-ABInflorescence-YInflorescence-WDFlower-FBFlower-FStamenCarpelPetalPollenRootLeaf-YLeaf-FSLeaf-SRachis-FSRachis-PFSRachis-VRachis-MRRachis-RSeedlingStem-GStem-WTendril-YTendril-WDTendril-FSClass IClass IIStemRootsLeaf - YLeaf - MEpithelial (28mm)Epithelial (41mm)Parenchyma (28mm)Parenchyma (41mm)Pre Anthesis (Ovule)Post Anthesis (Ovule)Fruit Set (Ovule)AlbedoFlavedoJuice SacsGO Description AtVvPtCsOsVvCsPtFGO:0050896 response to stimulus 5GO:0009628 response to abiotic stimulus 4GO:0042221 response to chemical stimulus 4GO:0010035 response to inorganic substance 3GO:0010033 response to organic substance 3GO:0009266 response to temperature stimulus 3GO:0016054 organic acid catabolic process 3GO:0006082 organic acid metabolic process 3GO:0043436 oxoacid metabolic process 3GO:0046395 carboxylic acid catabolic process 3GO:0019752 carboxylic acid metabolic process 3GO:0044281 small molecule metabolic process 3GO:0044282 small molecule catabolic process 3GO:0006573 valine metabolic process 3GO:0009081 branched-chain amino acid metabolism 3GO:0006814 sodium ion transport 2GO:0015672 monovalent inorganic cation transport 2BAC5 10 154 10 131                3                5Class I Class IIFigure 6 (See legend on next page.)Jia et al. BMC Plant Biology  (2015) 15:101 Page 16 of 23plawep wnour gvinentJia et al. BMC Plant Biology  (2015) 15:101 Page 17 of 23rice, poplar, grapevine and citrus (Additional file 11:Table S2-S9) were retrieved from publicly available co-expression databases [88-90]. In A. thaliana, the SDHClass I homologue (At5g51970) was significantly co-expressed with 67 genes (33% of total genes in the list)involved in branched chain amino acid metabolism,72 genes (36%) involved in response to various stim-uli, 37 genes (19%) involved in protein import in theperoxisome and 17 genes (9%) involved in auxin me-tabolism (Additional file 11: Table S2). In grapevines,the SDH Class I homologue (VIT_16s0100g00300)was significantly co-expressed with genes involved inabiotic stress (21%), peptide metabolism (13%) andlipid metabolism (13%) (Additional file 11: Table S3;Additional file 12: Table S2–S3). The co-expressionresults presented here corroborated with recent find-ings that the importance of SDH Class I lies in regu-lating sorbitol levels via its biochemical activity andgene expression during various abiotic stresses [83].More importantly, intracellular accumulation of sorbitolto high levels, accentuated under salt and osmotic stress,significantly reduced stress-induced biomass loss ofgrapevine berry cell suspensions which were likely theresults of the polyol utilisation as an effective osmopro-tectant and cellular homeostasis buffer [83]. Similar to itsArabidopsis counterpart (At5g51970), it is thereforelikely that grapevine SDH Class I plays an important rolein abiotic stress tolerance via the synergistic regulation ofpolyol transport and metabolism. The SDH Class IIhomologue (LIDH, VIT_16s0100g00290) was also signifi-(See figure on previous page.)Figure 6 Transcript and gene co-expression profiles of SDH in differenttissues and developmental stages of V. vinifera. Class I and II SDH genes50th percentile of all gene expression values, see Methods). The heatma(white) and 15 (red) to illustrate low, moderate and high expression wheClass I and Class II SDH gene in citrus. The heatmap was adjusted to colillustrate low, moderate and high expression when compared to all othe(adj. p-value) for genes co-expressed with SDHs from A. thaliana (At), V.frequencies in the plants tested. Light and dark orange denote enrichmcoloured in red. Grey colour denotes no significant enrichment.cantly co-expressed with genes related to abiotic stressresponse (35%). Other genes related to hexose biosyn-thetic pathways and carbohydrate metabolism (25%),protein biogenesis and catabolism (8%) and malic acidtransport (6%) were also evident in the list of co-expressed genes (Additional file 11: Table S4). GO termsassociated with these genes were also enriched within thegene lists (FDR < 0.05). Interestingly, GO enrichmentanalysis of co-expressed genes showed that terms as-sociated with “malate trans-membrane transport” and“response to abiotic stimulus” were highly enriched(FDR < 1.51E-04 and 3.5E–03 respectively) (Additionalfile 12: Table S2). Similarly to the grapevine SDH Class Igene, SDH Class II transcription was also stress respon-sive, being down-regulated during the heat stress recov-ery of grapevine leaves and up-regulated during exposureto UV-C light irradiation (Additional file 12: Table S3).Based on our coexpression analysis, we speculate that theinvolvement of Class II SDHs in abiotic stress responsesis likely to occur via a separate mechanism from that ofsorbitol metabolism, namely the ascorbate-glutathionecycle [91] and specifically in regulating the balance be-tween the biosynthesis of ascorbate by the L-galactosepathway [92] and its catabolism. This is supported in partin grapevines in which a marked down-regulation ofSDH Class II (LIDH) protein (impeding TA formation)and the up-regulation of proteins involved in L-galactosepathway (favouring Asc formation) in shoots of grape-vines during drought stress were observed [93]. There-fore, the stress responsive nature of SDH Class II geneand enzyme could potentially function as an extra levelof control (preventing loss of Asc to TA). The C. sinensisSDH Class II gene (Cit.9780.1.S1_at) was significantly co-expressed with genes involved in ion transport (11%),ubiquinone biosynthesis/oxidative phosphorylation (20%)and ribosome biogenesis (9%) (Additional file 11:Table S6). GO terms associated with these geneswere highly enriched within the co-expressed gene lists(Additional file 12: Table S5). Unlike Class I SDHs,enriched GO terms associated with Class II SDH co-expressed genes were more specialised to each corre-sponding plant but shared a common set of co-expressedgenes related to transporters (Additional file 11: Table S7;nts. A. Expression profiles for Class I and Class II SDH genes in variousre moderately to highly expressed in most tissues (Log2 intensity > 10;as adjusted to colour ranges between log2 intensity of 5 (blue), 10compared to all other genes respectively. B. Expression profiles forr ranges between log2 intensity of 4 (blue), 10 (white) and 14 (red) toenes respectively. C. Heatmap of selected enriched GO terms (−log10ifera (Vv), C. sinensis (Cs), P .trichocarpa [84], O. sativa (Os) and associatedscores between 1 and 3 respectively. Highly enriched scores (>5) areAdditional file 12: Table S6). In rice, the top 200 genesco-expressed with SDH (Os08g0545200) were primarilyenriched for genes involved in stress response (31%), car-boxylic acid biosynthesis (16%), plastid organisation(11%), protein transport (10%) and starch metabolism(5%) (Additional file 11: Table S5; Additional file 12:Table S4).Enriched GO parent terms such as “response tostimulus” and descendent terms “response to abioticstimulus”, were frequently enriched in SDH Class Ico-expressed lists and slightly in SDH Class II con-taining plant species (Figure 6C; Additional file 12:Table S1-S9). These observations agreed with previousJia et al. BMC Plant Biology  (2015) 15:101 Page 18 of 23reports that SDHs (Class I) in A. thaliana [13,23] andgrapevine [83] play an active role during drought stressand recovery processes and also suggest some sharedfunctions related to stress tolerance between the two clas-ses of SDH, even though to a conservative degree and po-tentially involving a separate mechanistic route. Therefore,enriched GO parent terms associated with “organic acidmetabolic process” and “branched-chain amino acid me-tabolism” were demonstrated to be more relevant to SDHClass I co-expressed genes but not to SDH Class II(Figure 6C). This is not surprising as response to variousstresses involves the coordinated regulation of amino acidand polyol accumulation [94]. On the other hand, co-expression analysis showed that plant SDH Class II couldbe tightly linked to mechanisms related to transport andcompartmentation of cations and solutes (Figure 6C). Inmembrane transport and compartmentation systems in-volving pumps, carriers and ion channels are also pivotalfor ion homeostasis and equivocally involved in a widerange of stress conditions [95]. In addition, divergent co-expression profiles across species have also been observedfor both classes of SDH. In general, monocot rice SDH-related genes have more common co-expression responseswith core Eudicot SDH Class I than with SDH Class II,corresponding with the finding that monocot SDH has acloser relationship with core Eudicot SDH Class I thanSDH Class II at the enzyme structural level.ConclusionsSDH is the key enzyme involved in sorbitol metabolismin higher plants. The results of the present study dem-onstrated that core Eudicot SDHs have evolved into twodistinct lineages: SDH Class I and SDH Class II. Class ISDH genes were present in all core Eudicot species in-vestigated in this study and appear to be essential for thenormal growth of plants. Class II SDH genes were foundto be absent in Brassicaceae, Leguminosae, most Asterids(except S. tuberosum) and some other plants. The previ-ously characterized LIDH involved in TA synthesis inV. vinifera has now been identified as a Class II SDHand represents a novel function of SDH genes in V. vi-nifera. The role of LIDH in TA synthesis may be rele-vant to the function of Class II SDHs in other species.Phylogeny, natural selection and genomic structure ana-lyses supported the emergence of SDH Class II as aresult of positive natural selection after tandem duplica-tion, which might occur in the common ancestor of coreEudicot plants. Furthermore, positive natural selectionhas only acted on specific amino acid sites in the SDHClass II lineage. Protein modelling analyses revealed sub-stitutions of three putative active site residues for Class Iand Class II SDHs, which may be responsible for theunique enzyme activity of V. vinifera LIDH. Gene ex-pression analysis demonstrated a clear transcriptionaldivergence between SDH Class I and Class II in severalplants and supports the divergence of Class II SDHs atthe expression level as well. Future work should be dedi-cated to uncovering the enzymatic activities and roles ofClass II SDH gene products in plant metabolism.MethodsIdentification of sorbitol dehydrogenase homologousgenes in higher plantsTo identify homologous SDHs in angiosperm plants, theamino acid sequence of A. thaliana SDH (accession no.At5g51970) was used as a query to BLAST against thegenomes of angiosperm species at Phytozome (http://www.phytozome.net/), with the exception of M. domes-tica for which genome dataset at Plant Genome Dupli-cation Database (PGDD, http://chibba.agtec.uga.edu/duplication/) was used instead. To increase datasetcoverage, the genomes of 8 recently sequenced speciesincluding Cajanus cajan, Jatropha curcas, Capsicumannuum, Brassica oleracea, Eutrema saisugineum,P. mume, Hordeum vulgare and Aegilops tauschii werealso queried using the corresponding genome databases.BLAST hits with an expectancy value (E value) of zerowere selected as SDH homologs were subjected to an-other round of BLAST searches within the genomesfrom which they were identified. Only the primary tran-script was chosen when alternative transcripts occurred.In addition, five partial SDH protein sequences ofP. bretschneideri [39] and one SDH sequence of Erio-botrya japonica [35] were obtained from literaturesearches. Homologous SDHs of P. hortorum wereprovided by the P. hortorum genome sequencing pro-ject author (Prof. Robert K. Jansen, The University ofTexas at Austin).Phylogenetic analysis of sorbitol dehydrogenaseThe Uniprot database was queried for previously identi-fied MDR mammal SDHs and yeast SDHs. Onlyreviewed entries were selected and used as the out-group in this phylogenetic analysis. Multiple sequencealignments of 102 sequences (92 plant SDHs, 7 mammalSDHs and 3 yeast SDHs) were carried out usingClustalW2 [96]. The evolutionary distances of targetSDHs (pairwise p-distance) were estimated using MEGA6software [97]. The Neighbour Joining tree was inferred byMEGA6 software [97] using the p-distance [98] substitu-tion model, the certainty at each node was assessed by theInterior-branch Test method (1000 times iteration). Max-imum likelihood trees were estimated by MEGA6 software[97] using the JTT+GAMMA substitution model [99], thebest fitting model as determined by the “Find Best DNA/Protein Models” function in MEGA6. Bootstrap supportsfor Maximum likelihood trees were calculated from 1000replicates. For both Neighbour Joining and MaximumJia et al. BMC Plant Biology  (2015) 15:101 Page 19 of 23likelihood methods, the Gaps/Missing Data Treatmentparameter was set as Complete-Deletion to eliminate theeffects of gaps and insertions. The developed phylogenetictrees were rooted on the yeast SDHs and annotated usingthe FigTree version 1.4.2 software (http://tree.bio.ed.ac.uk/software/figtree/).Sequence alignment and protein subdomain analysisPreliminary sequence identity of SDHs was obtained bylocal all-vs-all BLAST using NCBI-BLAST-2.2.29 tool[100] downloaded from ftp://ftp.ncbi.nlm.nih.gov/blast/executables/blast+/LATEST/. The BLAST results weresorted according to respective phylogeny groups. Aver-age pair-wise sequence identities were calculated usingMicrosoft Excel software based on the BLAST results.Protein functional domains were predicted usingInterPro (http://www.ebi.ac.uk/interpro/). Secondarystructure analysis was implemented with ESPript3.0 tool(http://espript.ibcp.fr/ESPript/ESPript/) using humanSDH (PDB: 1PL8) as a template. All residue numberingsin the present study are according to LIDH (Q1PSI9)without the first 20 amino acids (unless otherwise de-clared) which was predicted to be a mitochondria-targeting signal sequence (data not shown; alignmentcorresponding to this region was highly divergent).Gene duplication pattern characterization and syntenyanalysisThe MCScanX package [101] from http://chibba.pgml.uga.edu/mcscan2/ was employed to investigate gene du-plication patterns of plant SDHs. In order to elaborateon the origin of the core Eudicot Class II SDHs, plantgenomes containing SDHs from both Class I andClass II were selected. These were further refined to ge-nomes for which predicted genes have been mapped intocorresponding chromosome locations. A.thaliana wasincluded as a reference for inter-species collinear blockanalysis. Amino acid sequence files and gene positionfiles were downloaded either from PGDD or fromPhytozome databases and were further modified to suitthe requirements of the MCScanX software. BLAST toolNCBI-BLAST-2.2.29 [100] was used for intra and interspecies genome comparisons. The E-value threshold wasset at 10-5 for all analyses. For gene duplication patternidentification, self-genome all-vs-all BLAST was per-formed. The duplicate_gene_class ifier program fromthe MCScanX package was applied to each dataset.For collinear SDH gene pair identification, aminoacid sequences and genetic position information ofchromosomes containing SDHs were extracted fromeach species, then combined to perform the multi-species MCScanX analysis. The SDH gene family filewas created manually by including all the SDHs iden-tified from the selected species. The family_circle_plotter.java tool at MCScanX package was used to display theresults.Natural selection analysisNatural selective pressure on plant SDH was examinedby measuring the ratio of non-synonymous to synonym-ous substitutions (dN/dS=w). Codon-based maximum-likelihood estimates of w was performed using codemlin PAML4.7 [73]. Multiple-alignment of conserved do-main sequences (CDS) for those identified plant SDHswas carried out using ClustalW2 [96]. Significant inser-tions and gaps were removed manually. To facilitatethe input data requirements of codeml, an additionalMaximum Likelihood tree was constructed using asmaller dataset where SDHs with no CDS sequenceavailable were removed. The sub-tree covering the plantSDHs was used in codeml. Branch pattern specificationwas implemented using Treeview1.6.6 (http://taxonomy.zoology.gla.ac.uk/rod/treeview.html). Four target cladeswere specified based on the present phylogeneticanalysis: monocot SDH, A. coerulea SDH, core Eudi-cot SDH Class I and core Eudicot SDH Class II. Thew values for these clades were represented as w[mono], w[Aer], w[sdhC1] and w[sdhC2] respectively.Nested likelihood ratio tests(LRTs) were performed toassess the significance of the model under differenthypothesises: (w[mono]≠w[Aer]≠w[sdhC1]=w[sdhC2],w[mono]=w[Aer]≠w[sdhC1]≠w[sdhC2], w[mono]≠w[Aer]≠w[sdhC1]≠w[sdhC2], w[mono]=w[Aer]=w[sdhC1]≠w[sdhC2], w[mono]=w[Aer]=w[sdhC1]≠w[sdhC2] withw[sdhC2]=1). The corresponding p values were cal-culated using the online tool at http://graphpad.com/quickcalcs/PValue1.cfm. In the Site-specific model M1,two site classes were specified: highly conserved sites(w0) and neutral sites (w1=1). For the Site-specific modelM2, there were three site classes: highly conserved sites(w0), neutral sites (w1=1) and positively selected sites(w2). For w assessments with the Branch-site models,core Eudicot SDH Class II was specified as the fore-ground group. In the Branch-site model A, four site clas-ses were specified. The first two classes have w ratios ofw0 and w1 respectively, corresponding to highly con-served sites and neutral sites across all lineages. In theother two site classes, the background lineages have w0or w1 while the foreground lineages have w2.Ancestral sequence reconstruction and evolution rateanalysesThe ancestral sequence (amino acid) reconstruction forthe internal nodes of the obtained plant SDH phylogenywas carried out using codeml in PAML4.7 [73]. TheEmpirical_Frequency model, which allowed the esti-mates of the stationary frequencies based on user data-set, was performed on the plant SDHs. Ancestral aminoJia et al. BMC Plant Biology  (2015) 15:101 Page 20 of 23acid sequences for nodes representing monocot SDH,A. coerulea SDH, core Eudicot SDH Class I and coreEudicot SDH Class II were used for Tajima’s RRT ana-lysis [74] using MEGA6.0 software [97].Protein structure modelling analysisSDH homology modelling was carried out using ICMPro (Molsoft LLC, La Jolla, CA, USA). Models of V. vi-nifera LIDH (Uniprot ID: Q1PSI9; accession no:GSVIVT01010646001) and V. vinifera SDH (Uniprot ID:D7TMY3; accession no: GSVIVT01010642001) struc-tures were generated with the human SDH (PDB:1PL8)as a template. Given that no plant SDH structures existin the protein data bank we chose the model with thehighest identity as performed within the Molsoft soft-ware package. Ligands including the zinc atom, NAD+,D-sorbitol and L-idonate were docked into the modelsusing the Molsoft Monte Carlo method [102]. Residueswithin 5 Å to the ligands were inspected for enzyme-ligand interaction potential. All molecular visualiza-tions were obtained using the PyMOL graphic tool(The PyMOL molecular graphics system, Version 1.3r1.Schrodinger, LLC). The deduced reaction mechanismof V. vinifera LIDH on the oxidation of L-idonatewas created using the Marvin online tool (http://www.chemaxon.com/marvin/sketch/index.php). Pro-tein hydrophobicity profiles were implemented inPyMOL using the Color_h script (http://www.pymolwiki.org/index.php/Color_h), based on the hydrophobicityscale defined at http://us.expasy.org/tools/pscale/Hphob.Eisenberg.html. All residue numberings are according toLIDH (Q1PSI9) without the first 20 amino acids.Meta-analysis of developmental gene expressionIdentification of corresponding probesets in the micro-array platforms of A. thaliana, rice, poplar, grapevineand citrus were performed using the BLAST software(NCBI-BLAST-2.2.29+) [100], and grapevine Class I(VIT_16s0100g00290) and Class II (VIT_16s0100g00290)SDH sequences with default settings. The top hits foreach corresponding probeset in the microarray platformof each species were selected for downstream analysis(Additional file 11). Normalised gene expression at-lases encompassing transcriptional data during growthand development of A. thaliana, grapevine and citruswere retrieved from the Botany Array Resource (BAR)[80], Vitis co-expression database (VTCdb) [88] andNetwork inference of citrus co-expression (NiCCE)[89] webservers, respectively. Only experimental con-ditions relating to tissue/organ development and pro-besets intensities (normalised) corresponding to Class Iand Class II SDHs were retained. Normalised log2intensities were deemed highly, well and lowly/notexpressed when the intensities of total backgrounddistribution > 95th, at the 50th and < 20th percentilerespectively.Gene co-expression mining in various plant speciesInformation on co-expressed genes with Class I andClass II SDHs in plants such as A. thaliana, poplar andrice (version 7.1) [90], grapevine (version 2.1) [88] andcitrus [89] were retrieved from the various plant geneco-expression webservers. The top 200 co-expressedgenes (unless otherwise specified) for each SDH class ineach species were empirically chosen as a cut-off for sig-nificant co-expression, and to provide comparisons ofenriched gene ontology (GO) terms within the co-expressed gene lists from each species. Enrichment ofGO terms (i.e. biological processes, BP; molecular function,MF; cellular component, CC) were evaluated by hypergeo-metric distribution, adjusted by false discovery rate (FDR)for multiple hypothesis correction and using the ‘gProfileR’package [103] in R (http://www.r-project.org) which inter-faces g:profiler webserver (http://gprofiler.at.mt.ut.ee/gprofiler/). The ‘ordered query’ option was enabled toperform incremental enrichment analysis, which priori-tises highly co-expressed genes and results in better func-tional GO term associations. GO terms were consideredto be significantly enriched when FDR < 0.05 and > 2genes were annotated with the same GO term. EnrichedGO terms from the SDH co-expressed gene lists acrosstested plants (A. thaliana, poplar, rice, grapevine and cit-rus), were considered ‘commonly occurring’ when morethan 3 counts were present for each enriched GO term.Availability of supporting dataAll relevant supporting data can be found within theadditional files accompanying this article. Phylogeneticdata supporting the results of this article are availablein the TreeBASE repository at http://purl.org/phylo/treebase/phylows/study/TB2:S17300.Additional filesAdditional file 1: Displays the molecular structures of SDH substrates.Additional file 2: Table S1. Contains SDH gene IDs from correspondingspecies and organisms. Table S2. Contains pairwise p-distance values ofSDH sequences. Table S3. Contains information on sequence renaming.Table S4. Contains the all-vs-all BLAST results of SDH amino acid sequences.Table S5 contain the identified collinear SDH gene pairs.Additional file 3: Contains the original amino acid sequences of theidentified plant, mammal and yeast SDHs.Additional file 4: Displays the complete Neighbour Joining tree forFigure 2A.Additional file 5: Displays complete sequence alignment for Figure 3.Additional file 6: Contains gene duplication pattern information.Tables “cs”, “eg”, “md”, “pm”, “pp”, “pt”, “st”, “tc”, “vv” refer to C. sinensis,E. grandis, M. domestica, P. mume, P. persica, P. trichocarpa, S. tuberosum,T. cacao and V. vinifera respectively.by fructose and auxin. J Jpn Soc Hortic Sci. 2008;77(3):318–23.Jia et al. BMC Plant Biology  (2015) 15:101 Page 21 of 23Additional file 7: Contains input and output data for naturalselection modelling analyses. “-output” files are codeml outputs andare recommended to be viewed using Microsoft WordPad. “.phy” isphylogenetic tree file and can be viewed using Treeview software. “.ctl” isa control file and can be viewed using any text viewer. “sdh-pep2.fas”sequence file was produced from Additional file 3 by manually removingthe significant gaps, insertions; sequences with no CDS sequence availablewere also removed. “sdh-cds2.fas” is the corresponding CDS sequences for“sdh-pep2.fas”. “sdh-pep2.nwk” is the phylogenetic tree produced from“sdh-cds2.fas” and can be viewed using any phylogenetic tree viewersoftware. Sequence IDs are represented by numbers for software inputconvenience (see Additional file 2: Table S3 for sequence ID renaminginformation). Amino acid site numbering is according to LIDH (UniprotNo: Q1PSI9) without the first 20 amino acids.Additional file 8: Contains input and output data for thereconstruction of ancestral SDH sequences. “sdh-pep.fas” containsamino acid sequences for the plant SDH sub-branch. The “ancestral-sequence-construction_output” file is codeml output and can beviewed using any text viewer. Ancestral sequences for correspondingbranches were extracted and put in the “interpreted-ancestral-sequence.fas”file for readers’ convenience.Additional file 9: Contains the Tajima’s RRT test outputs.Additional file 10: Contains the modelled structures files of Vv_LIDHand Vv_SDH and additional illustration figures. “Asp195_NAD.png”displays the interaction of Asp195 with the hydroxyl groups at C1 and C2 ofL-idonate. “LIDH-hydrophobicity.png” and “SDH-hydrophobicity.png” displaythe overall hydrophobicity profiles of Vv_LIDH and Vv_SDH respectively.Amino acid site numbering is according to LIDH (Uniprot No: Q1PSI9)without the first 20 amino acids.Additional file 11: Contains a Microsoft Excel spread sheet withdetailed results of transcript and gene co-expression analysis ofClass I and Class II SDH in plants. Table S1 contains gene expressionprofile of Class I and Class II SDH profile in various tissues of (A) grapevineand (B) sweet oranges. Table S2 – S9 contains lists of all significantlyco-expressed genes and respective rank, function description, andco-expression metric with class I and II SDH in A. thaliana (Table S2),grapevine (Table S3 and S4), rice (Table S5), sweet orange (Table S6and S7) and poplar (Table S8 and S9).Additional file 12: Contains a Microsoft Excel spread sheet withdetailed results of functional (GO) enrichment analysis ofsignificantly co-expressesed genes of class I and II SDH in plants.Table S1 – S8 contains outputs of GO enrichment analysis containingenriched GO ID, description, adjusted p-value, and lists of genes havingthe enriched GO term for A. thaliana (Table S1), grapevine (Table S2 and S3),rice (Table S4), sweet orange (Table S5 and S6) and poplar (Table S7 and S8).Table S9 contains a summary of common enriched GO ID/term identifiedamong the co-expressed genes with SDHs in the aforementionedplants tested.AbbreviationsSDH: Sorbitol dehydrogenase; LIDH: L-idonate-5-dehydrogenase; TA: Tartaricacid; MDR: Medium-chain dehydrogenase/reductase; ADH: Alcoholdehydrogenase; 5KGA: 5-keto-D-gluconate; Mya: Million years ago; WGD: Wholegenome duplication; RRT: Relative rate tests; GCA: Gene co-expression networkanalysis; PGDD: Plant Genome Duplication Database; CDS: Conserved domainsequences; LRTs: Likelihood ratio tests; BAR: Botany Array Resource; VTCdb: Vitisco-expression database; NiCCE: Network inference of citrus co-expression;GO: Gene ontology; BP: Biological processes; Asc: Ascorbate; MF: Molecularfunction; CC: Cellular component; FDR: False discovery rate.Competing interestsThe authors declare that they have no competing interests.Authors’ contributionsYJ conceived the research. YJ and DCJW did sequence retrieval, curationand gene duplication characterization. YJ performed phylogenetic, synteny,natural selection modeling and ancestral sequence analyses and draftedthe manuscript. JBB and YJ carried out protein modeling analyses. DCJWperformed the transcript expression and gene co-expression analysis. CS13. Aquayo MF, Ampuero D, Mandujano P, Parada R, Muñoz R, Gallart M, et al.Sorbitol dehydrogenase is a cytosolic protein required for sorbitolmetabolism in Arabidopsis thaliana. Plant Sci. 2013;205–206(1):63–75.14. Persson B, Zigler JS, Jornvall H. A super-family of medium-chainand DCJW assisted with the drafting of the manuscript. CMF and JBBsupervised the project. All authors have read and approved the finalmanuscript.AcknowledgementsWe acknowledge the related research groups for making the genomicinformation and microarray data available to the public. We are very gratefulto Dr Anthony Borneman and Dr Julian Schwerdt for their valuablesuggestions about the phylogenetic analyses. We thank the anonymousreferees for their constructive comments and suggestions. This work waspart-supported by Australia’s grape growers and winemakers through theGrape and Wine Research and Development Corporation with matchingfunds from the Australian Government (project UA 10/01). YJ is supported bya postgraduate scholarship from China Scholarship Council.Author details1School of Agriculture, Food and Wine, University of Adelaide, Adelaide 5005,Australia. 2Present address: Wine Research Center, Faculty of Land and FoodSystems, University of British Columbia, Vancouver V6T 1Z4BC, Canada.3Present address: School of Biological Sciences, Flinders University, GPO Box2100, Adelaide 5001, Australia. 4School of Biological Sciences, University ofAdelaide, Adelaide 5005, Australia.Received: 14 November 2014 Accepted: 23 March 2015References1. Iwata T, Hoog JO, Reddy VN, Carper D. Cloning of the human SorbitolDehydrogenase gene. Invest Ophth Vis Sci. 1993;34(4):712–2.2. Karlsson C, Jornvall H, Hoog JO. Sorbitol Dehydrogenase - cDNA coding forthe rat enzyme - variations within the Alcohol-Dehydrogenase familyindependent of quaternary structure and metal content. Eur J Biochem.1991;198(3):761–5.3. Wang T, Hou M, Zhao N, Chen Y, Lv Y, Li Z, et al. Cloning and expression ofthe sorbitol dehydrogenase gene during embryonic development andtemperature stress in Artemia sinica. Gene. 2013;521(2):296–302.4. Niimi T, Yamashita O, Yaginuma T. A cold-inducible Bombyx geneencoding a protein similar to mammalian Sorbitol Dehydrogenase - yolknuclei-dependent gene-expression in diapause eggs. Eur J Biochem.1993;213(3):1125–31.5. Sarthy AV, Schopp C, Idler KB. Cloning and sequence determination of thegene encoding Sorbitol Dehydrogenase from Saccharomyces cerevisiae.Gene. 1994;140(1):121–6.6. Ng K, Ye RQ, Wu XC, Wong SL. Sorbitol Dehydrogenase from Bacillussubtilis - purification, characterization, and gene cloning. J Biol Chem.1992;267(35):24989–94.7. Negm FB, Loescher WH. Detection and characterization of SorbitolDehydrogenase from apple callus-tissue. Plant Physiol. 1979;64(1):69–73.8. Doehlert DC. Ketose reductase-activity in developing maize endosperm.Plant Physiol. 1987;84(3):830–4.9. Yamada K, Oura Y, Mori H, Yamaki S. Cloning of NAD-dependent sorbitoldehydrogenase from apple fruit and gene expression. Plant Cell Physiol.1998;39(12):1375–9.10. Park SW, Song KJ, Kim MY, Hwang JH, Shin YU, Kim WC, et al. Molecularcloning and characterization of four cDNAs encoding the isoforms ofNAD-dependent sorbitol dehydrogenase from the Fuji apple. Plant Sci.2002;162(4):513–9.11. Ohta K, Moriguchi R, Kanahama K, Yamaki S, Kanayama Y. Molecularevidence of sorbitol dehydrogenase in tomato, a non-Rosaceae plant.Phytochemistry. 2005;66(24):2822–8.12. Sutsawat D, Yamada K, Shiratake K, Kanayama Y, Yamaki S. Properties ofsorbitol dehydrogenase in strawberry fruit and enhancement of the activitydehydrogenases/reductases (MDR) - Sub-lines including zeta-crystallin,alcohol and polyol dehydrogenases, quinone oxidoreductases, enoylreductases, Vat-1 and other proteins. Eur J Biochem. 1994;226(1):15–22.Jia et al. BMC Plant Biology  (2015) 15:101 Page 22 of 2315. Persson B, Hedlund J, Jornvall H. The MDR superfamily. Cell Mol Life Sci.2008;65(24):3879–94.16. Jornvall H, Persson M, Jeffery J. Alcohol and polyol dehydrogenases areboth divided into 2 protein types, and structural-properties cross-relate thedifferent enzyme-activities within each type. P Natl Acad Sci-Biol.1981;78(7):4226–30.17. Nordling E, Jornvall H, Persson B. Medium-chain dehydrogenases/reductases(MDR) - family characterizations including genome comparisons and activesite modelling. Eur J Biochem. 2002;269(17):4267–76.18. Lindstad RI, Koll P, McKinley-McKee JS. Substrate specificity of sheep liversorbitol dehydrogenase. J Biochemical. 1998;330(Pt 1):479–87.19. Lindstad RI, Hermansen LF, McKinley-Mckee JS. The kinetic mechanism ofsheep liver sorbitol dehydrogenase. Eur J Biochem. 1992;210(2):641–7.20. Oura Y, Yamada K, Shiratake K, Yamaki S. Purification and characterization ofa NAD(+)-dependent sorbitol dehydrogenase from Japanese pear fruit.Phytochemistry. 2000;54(6):567–72.21. Guo ZX, Pan TF, Li KT, Zhong FL, Lin L, Pan DM, et al. Cloning of NAD-SDHcDNA from plum fruit and its expression and characterization. Plant PhysiolBiochem. 2012;57:175–80.22. Pauly TA, Ekstrom JL, Beebe DA, Chrunyk B, Cunningham D, Griffor M, et al.X-ray crystallographic and kinetic studies of human sorbitol dehydrogenase.Structure. 2003;11(9):1071–85.23. Nosarzewski M, Downie AB, Wu B, Archbold DD. The role of sorbitoldehydrogenase in Arabidopsis thaliana. Funct Plant Biol. 2012;39(6):462–70.24. Yancey PH, Clark ME, Hand SC, Bowlus RD, Somero GN. Living with water-stress - evolution of osmolyte systems. Science. 1982;217(4566):1214–22.25. Loescher WH. Physiology and metabolism of sugar alcohols in higher-plants. Physiol Plantarum. 1987;70(3):553–7.26. Loescher WH, Marlow GC, Kennedy RA. Sorbitol metabolism and sink-sourceinterconversions in developing apple leaves. Plant Physiol. 1982;70(2):335–9.27. Nosarszewski M, Clements AM, Downie AB, Archbold DD. Sorbitoldehydrogenase expression and activity during apple fruit set and earlydevelopment. Physiol Plantarum. 2004;121(3):391–8.28. Nosarzewski M, Archbold DD. Tissue-specific expression of sorbitoldehydrogenase in apple fruit during early development. J Exp Bot.2007;58(7):1863–72.29. Wang XL, Xu YH, Peng CC, Fan RC, Gao XQ. Ubiquitous distribution anddifferent subcellular localization of sorbitol dehydrogenase in fruit and leafof apple. J Exp Bot. 2009;60(3):1025–34.30. Yamaguchi H, Kanayama Y, Soejima J, Yamaki S. Changes in the amounts ofthe NAD-dependent sorbitol dehydrogenase and its involvement in thedevelopment of apple fruit. J Am Soc Hortic Sci. 1996;121(5):848–52.31. Wu BH, Li SH, Nosarzewski M, Archbold DD. Sorbitol dehydrogenase geneexpression and enzyme activity in apple: tissue specificity during buddevelopment and response to rootstock vigor and growth manipulation.J Am Soc Hortic Sci. 2010;135(4):379–87.32. Iida M, Bantog NA, Yamada K, Shiratake K, Yamaki S. Sorbitol- and othersugar-induced expressions of the NAD+−dependent sorbitol dehydrogenasegene in Japanese pear fruit. J Am Soc Hortic Sci. 2004;129(6):870–5.33. Kim HY, Ahn JC, Choi JH, Hwang B, Choi DW. Expression and cloningof the full-length cDNA for sorbitol-6-phosphate dehydrogenase andNAD-dependent sorbitol dehydrogenase from pear (Pyrus pyrifolia N.).Sci Hortic. 2007;11(4):406–12.34. Bantog NA, Shiratake K, Yamaki S. Changes in sugar content and sorbitol-and sucrose-related enzyme activities during development of loquat(Eriobotrya japonica Lindl. cv. Mogi) fruit. J Jpn Soc Hortic Sci. 1999;68(5):942–8.35. Bantog NA, Yamada K, Niwa N, Shiratake K, Yamaki S. Gene expressionof NAD(+)-dependent sorbitol dehydrogenase and NADP(+)-dependentsorbitol-6-phosphate dehydrogenase during development of loquat(Eriobotrya japonica Lindl.) fruit. J Jpn Soc Hortic Sci. 2000;69(3):231–6.36. Beruter J. Sugar accumulation and changes in the activities of relatedenzymes during development of the apple fruit. J Plant Physiol.1985;121(4):331–41.37. Kuo TM, Doehlert DC, Crawford CG. Sugar metabolism in germinatingsoybean seeds - evidence for the sorbitol pathway in soybean axes. PlantPhysiol. 1990;93(4):1514–20.38. de Sousa SM, Paniago MD, Arruda P, Yunes JA. Sugar levels modulate sorbitoldehydrogenase expression in maize. Plant Mol Biol. 2008;68(3):203–13.39. Ito A, Hayama H, Kashimura Y. Partial cloning and expression analysis ofgenes encoding NAD(+)-dependent sorbitol dehydrogenase in pear budduring flower bud formation. Sci Hortic. 2005;103(4):413–20.40. Hartman MD, Figueroa CM, Piattoni CV, Iglesias AA. Glucitol Dehydrogenasefrom peach (Prunus persica) fruits is regulated by thioredoxin h. Plant CellPhysiol. 2014;55(6):1157–68.41. Flagel LE, Wendel JF. Gene duplication and evolutionary novelty in plants.New Phytol. 2009;183(3):557–64.42. Zhang JZ. Evolution by gene duplication: an update. Trends Ecol Evol.2003;18(6):292–8.43. Hughes AL. The evolution of functionally novel proteins after geneduplication. P Roy Soc B-Biol Sci. 1994;256(1346):119–24.44. Hughes AL. Adaptive evolution after gene duplication. Trends Genet.2002;18(9):433–4.45. Hurles M. Gene duplication: the genomic trade in spare parts. Plos Biol.2004;2(7):900–4.46. Force A, Lynch M, Pickett FB, Amores A, Yan YL, Postlethwait J. Preservationof duplicate genes by complementary, degenerative mutations. Genetics.1999;151(4):1531–45.47. Conant GC, Wolfe KH. Turning a hobby into a job: how duplicated genesfind new functions. Nat Rev Genet. 2008;9(12):938–50.48. DeBolt S, Cook DR, Ford CM. L-Tartaric acid synthesis from vitamin C inhigher plants. P Natl Acad Sci USA. 2006;103(14):5608–13.49. Strommer J. The plant ADH gene family. Plant J. 2011;66(1):128–42.50. Velasco R, Zharkikh A, Affourtit J, Dhingra A, Cestaro A, Kalyanaraman A,et al. The genome of the domesticated apple (Malus x domestica Borkh.).Nat Genet. 2010;42(10):833–9.51. Forney CF, Breen PJ. Growth of strawberry fruit and sugar uptake of fruitdisks at different inflorescence positions. Sci Hortic. 1985;27(1–2):55–62.52. Veitia RA, Bottani S, Birchler JA. Cellular reactions to gene dosageimbalance: genomic, transcriptomic and proteomic effects. Trends Genet.2008;24(8):390–7.53. Galdon BR, Mesa DR, Rodriguez EMR, Romero CD. Influence of the cultivaron the organic acid and sugar composition of potatoes. J Sci Food Agric.2010;90(13):2301–9.54. Stafford HA. Distribution of tartaric acid in the leaves of certain angiosperms.Am J Bot. 1959;46(5):347–52.55. Kramer EM. Aquilegia: a new model for plant development, ecology, andevolution. Annu Rev Plant Biol. 2009;60:261–77.56. Worberg A, Quandt D, Barniske AM, Lohne C, Hilu KW, Borsch T. Phylogenyof basal eudicots: insights from non-coding and rapidly evolving DNA. OrgDivers Evol. 2007;7(1):55–77.57. Hoot SB, Magallon S, Crane PR. Phylogeny of basal eudicots based on threemolecular data sets: atpB, rbcL, and 18S nuclear ribosomal DNA sequences.Ann Mo Bot Gard. 1999;86(1):1–32.58. Moore MJ, Bell CD, Soltis PS, Soltis DE. Using plastid genome-scale data toresolve enigmatic relationships among basal angiosperms. P Natl Acad SciUSA. 2007;104(49):19363–8.59. Wang HC, Moore MJ, Soltis PS, Bell CD, Brockington SF, Alexandre R, et al.Rosid radiation and the rapid rise of angiosperm-dominated forests. P NatlAcad Sci USA. 2009;106(10):3853–8.60. Bremer B, Bremer K, Chase MW, Fay MF, Reveal JL, Soltis DE, et al. Anupdate of the angiosperm phylogeny group classification for the ordersand families of flowering plants: APG III. Bot J Linn Soc. 2009;161(2):105–21.61. Roulin A, Auer PL, Libault M, Schlueter J, Farmer A, May G, et al. The fate ofduplicated genes in a polyploid plant genome. Plant J. 2013;73(1):143–53.62. Saito K, Loewus FA. Formation of tartaric acid in Vitaceous plants - relativecontributions of L-ascorbic acid-inclusive and acid-noninclusive pathways.Plant Cell Physiol. 1989;30(6):905–10.63. Nour V, Trandafir I, Ionica ME. HPLC organic acid analysis in differentcitrus juices under reversed phase conditions. Not Bot Horti Agrobo.2010;38(1):44–8.64. Hudina M, Stampar F. Sugars and organic acids contents of European(Pyrus communis L.) and Asian (Pyrus serotina Rehd.) pear cultivars. ActaAliment Hung. 2000;29(3):217–30.65. Sha SF, Li JC, Wu J, Zhang SL. Characteristics of organic acids in the fruit ofdifferent pear species. Afr J Agr Res. 2011;6(10):2403–10.66. Fuleki T, Pelayo E, Palabay RB. Carboxylic-acid composition of varietal juicesproduced from fresh and stored apples. J Agr Food Chem. 1995;43(3):598–607.67. Suarez MH, Rodriguez ER, Romero CD. Analysis of organic acid content incultivars of tomato harvested in Tenerife. Eur Food Res Technol.2008;226(3):423–35.68. Ina Y. Pattern of synonymous and nonsynonymous substitutions: an indicatorof mechanisms of molecular evolution. J Genet. 1996;75(1):91–115.evolutionary genetics analysis Version 6.0. Mol Biol Evol. 2013;30(12):2725–9.98. Nei M, Kumar S. Molecular Evolution and Phylogenetics. New York: OxfordUniversity Press; 2000.99. Jones DT, Taylor WR, Thornton JM. The rapid generation of mutation datamatrices from protein sequences. Comput Appl Biosci. 1992;8(3):275–82.100. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignmentsearch tool. J Mol Biol. 1990;215(3):403–10.101. Wang YP, Tang HB, DeBarry JD, Tan X, Li JP, Wang XY, et al. MCScanX: atoolkit for detection and evolutionary analysis of gene synteny andcollinearity. Nucleic Acids Res. 2012;40(7):e49.102. Abagyan R, Totrov M. Biased probability Monte-Carlo conformationalsearches and electrostatic calculations for peptides and proteins. J Mol Biol.1994;235(3):983–1002.103. Reimand J, Arak T, Vilo J. g: profiler-a web server for functional interpretationof gene lists (2011 update). Nucleic Acids Res. 2011;39(suppl2):W307–15.Jia et al. BMC Plant Biology  (2015) 15:101 Page 23 of 2369. Kimura M. Preponderance of synonymous changes as evidence for the neutraltheory of molecular evolution. Letters to Nature. 1977;267(5608):275–6.70. Yang ZH. PAML: a program package for phylogenetic analysis by maximumlikelihood. Comput Appl Biosci. 1997;13(5):555–6.71. Worth CL, Gong S, Blundell TL. Structural and functional constraints in theevolution of protein families. Nat Rev Mol Cell Bio. 2009;10(10):709–20.72. Yang ZH, Nielsen R. Codon-substitution models for detecting molecularadaptation at individual sites along specific lineages. Mol Biol Evol.2002;19(6):908–17.73. Yang ZH. PAML 4: phylogenetic analysis by maximum likelihood. Mol BiolEvol. 2007;24(8):1586–91.74. Tajima F. Simple methods for testing the molecular evolutionary clockhypothesis. Genetics. 1993;135(2):599–607.75. Baker PJ, Britton KL, Rice DW, Rob A, Stillman TJ. Structural consequencesof sequence patterns in the fingerprint region of the nucleotide bindingfold - implications for nucleotide specificity. J Mol Biol. 1992;228(2):662–71.76. Yamaguchi H, Kanayama Y, Yamaki S. Purification and properties of NAD-dependent Sorbitol Dehydrogenase from apple fruit. Plant Cell Physiol.1994;35(6):887–92.77. Yennawar H, Moller M, Gillilan R, Yennawar N. X-ray crystal structure andsmall-angle X-ray scattering of sheep liver sorbitol dehydrogenase. ActaCrystallogr D. 2011;67(Pt5):440–6.78. Gu ZL, Nicolae D, Lu HHS, Li WH. Rapid divergence in expressionbetween duplicate genes inferred from microarray data. Trends Genet.2002;18(12):609–13.79. Wagner A. Decoupled evolution of coding region and mRNA expressionpatterns after gene duplication: implications for the neutralist-selectionistdebate. P Natl Acad Sci USA. 2000;97(12):6579–84.80. Toufighi K, Brady SM, Austin R, Ly E, Provart NJ. The botany array resource:e-northerns, expression angling, and promoter analyses. Plant J.2005;43(1):153–63.81. Fasoli M, Dal Santo S, Zenoni S, Tornielli GB, Farina L, Zamboni A,et al. The grapevine expression atlas reveals a deep transcriptomeshift driving the entire plant into a maturation program. Plant Cell.2012;24(9):3489–505.82. Sweetman C, Wong DCJ, Ford CM, Drew DP. Transcriptome analysis at fourdevelopmental stages of grape berry (Vitis vinifera cv. Shiraz) providesinsights into regulated and coordinated gene expression. BMC Genomics.2012;13(1):691.83. Conde A, Regalado A, Rodrigues D, Costa JM, Blumwald E, Chaves MM,et al. Polyols in grape berry: transport and metabolic adjustments as aphysiological strategy for water-deficit stress tolerance in grapevine. J ExpBot. 2015;66(3):889–906.84. Mazzitelli L, Hancock RD, Haupt S, Walker PG, Pont SDA, McNicol J, et al.Co-ordinated gene expression during phases of dormancy release inraspberry (Rubus idaeus L.) buds. J Exp Bot. 2007;58(5):1035–45.85. Pilati S, Perazzolli M, Malossini A, Cestaro A, Dematte L, Fontana P, et al.Genome-wide transcriptional analysis of grapevine berry ripening reveals aset of genes similarly modulated during three seasons and the occurrenceof an oxidative burst at veraison. BMC Genomics. 2007;8(1):428.86. Fortes AM, Agudelo-Romero P, Silva MS, Ali K, Sousa L, Maltese F,et al. Transcript and metabolite analysis in Trincadeira cultivar revealsnovel information regarding the dynamics of grape ripening. BMCPlant Biol. 2011;11(1):149.87. Melino VJ, Soole KL, Ford CM. A method for determination of fruit-derivedascorbic, tartaric, oxalic and malic acids, and its application to thestudy of ascorbic acid catabolism in grapevines. Aust J Grape WineRes. 2009;15(3):293–302.88. Wong DCJ, Sweetman C, Drew DP, Ford CM. VTCdb: a gene co-expressiondatabase for the crop species Vitis vinifera (grapevine). BMC Genomics.2013;14(1):17.89. Wong DCJ, Sweetman C, Ford CM. Annotation of gene function in citrususing gene expression information and co-expression networks. BMC PlantBiol. 2014;14(1):17.90. Obayashi T, Okamura Y, Ito S, Tadaka S, Aoki Y, Shirota M, et al. ATTED-II in2014: evaluation of gene coexpression in agriculturally important plants.Plant Cell Physiol. 2014;55(1):e6.91. Foyer CH, Noctor G. Ascorbate and glutathione: the heart of the redox hub.Plant Physiol. 2011;155(1):2–18.92. Wheeler GL, Jones MA, Smirnoff N. The biosynthetic pathway of vitamin Cin higher plants. Nature. 1998;393(6683):365–9.Submit your next manuscript to BioMed Centraland take full advantage of: • Convenient online submission• Thorough peer review• No space constraints or color figure charges• Immediate publication on acceptance• Inclusion in PubMed, CAS, Scopus and Google Scholar• Research which is freely available for redistribution93. Cramer GR, Van Sluyter SC, Hopper DW, Pascovici D, Keighley T, Haynes PA.Proteomic analysis indicates massive changes in metabolism prior to theinhibition of growth and photosynthesis of grapevine (Vitis vinifera L.) inresponse to water deficit. BMC Plant Biol. 2013;13(1):49.94. Krasensky J, Jonak C. Drought, salt, and temperature stress-inducedmetabolic rearrangements and regulatory networks. J Exp Bot.2012;63(4):1593–608.95. Conde A, Chaves MM, Geros H. Membrane transport, sensing andsignaling in plant adaptation to environmental stress. Plant CellPhysiol. 2011;52(9):1583–602.96. Larkin MA, Blackshields G, Brown NP, Chenna R, McGettigan PA, McWilliam H,et al. Clustal W and Clustal X version 2.0. Bioinformatics. 2007;23(21):2947–8.97. Tamura K, Stecher G, Peterson D, Filipski A, Kumar S. MEGA6: molecularSubmit your manuscript at www.biomedcentral.com/submit


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items