Knowledge of the GC content of a genome and codon usage is useful when devising PCR-based strategies for mapping and gene isolation, as well as for hybridization studies by microarray. Based on the coding regions of 302 A. deliciosa, 319 A. chinensis and 84 A. eriantha full-length cDNAs, the GC content in the third base position was estimated to be 55%, 49% and 58%, respectively. These values are higher than the overall GC ratio from the sequences of the complete EST datasets of A. deliciosa (46%), A. chinensis (46%) and A. eriantha (48%), indicating that there is some pressure, particularly in A. deliciosa and A. eriantha, towards an increased GC ratio in coding regions compared with non-translated regions. Overall the codon usage of the three Actinidia species is similar, although not identical (Additional file 3). A. deliciosa and A. eriantha differ in only their preference for aspartate (GAT/GAC) and serine (TCT/TCC). In contrast, comparative codon preferences between A. deliciosa and A. chinensis show differences for alanine (GCC/GCT), glycine (GGC/GGA), isoleucine (ATC/ATT), leucine (CTC/TTG), asparagine (AAC/AAT), glutamine (CAG/CAA), threonine (ACC/ACT) and valine (GTG/GTT). In most of these codons, A. chinensis shows a greater preference for an A or T in the third base position, accounting for its lower GC third base percentage when compared with the other two species.

Both SNPs and microsatellites or SSRs are valuable tools for genetic mapping within breeding populations of many crops. A total of 32,764 biallelic SNPs were detected from the overlapping regions of 3,901 (21.6%) of the 18,070 TCs, at a rate of one SNP every 417 bp. This frequency is higher than that reported in apple using a dataset of comparable size (one SNP per 706 bp from ~150,000 ESTs 10). As several Actinidia species were used to construct the cDNA libraries, while the apple cDNA libraries were from different varieties of the same species, this increased frequency of SNPs could reflect the greater genetic diversity in the surveyed Actinidia EST libraries. For example the frequency in apple increased significantly to 1 in 149 bp when a greater diversity of genotypes was added (Chagne et al unpublished results). The polyploidy nature of several of the Actinidia species used may also have affected SNP frequencies. As a result of this, some of the SNPs identified may not be allelic in nature but due to homoeologous or paralogous sequences clustering in the same TC.

EST-derived SSRs have already been shown to be valuable mapping tools in Actinidia chinensis 14 where 150 SSRs with more than 10 dinucleotide repeat units were tested as markers in an A. chinensis mapping population. More than 90% of the SSR markers were polymorphic and segregated within the population. Subsequently, 20 of these SSRs were shown to be transportable across multiple species of Actinidia, showing the value of this resource 15.

Compared with apple 10, where less than 20% of the NRs contained a microsatellite, over 30% of Actinidia NRs were found to have at least one. The other major difference between apple and Actinidia was that, while di- and tri-nucleotide repeats were equally frequent in apple at 7 to 8% frequency, di-nucleotide repeats (18%) were twice as frequent as tri-nucleotide repeats (9%) in Actinidia. Tetra-nucleotide repeats were found at similar frequencies in NRs for the two genera. The lengths of the SSRs were similar in both apple and Actinidia, with 50% of the repeats having between 12 and 14 bases. As in apple, AG repeats were the most frequent (16% of NRs, significantly more frequent than in apple) and were followed by AT (1.2%) and AC (0.8%) while GC repeats were very infrequent.

The position of SSRs in relation to the putative initiation ATG was very similar when comparing Actinidia with other plants such as Arabidopsis and apple 1020, with di-nucleotide repeats being predominantly (93%) in the 5' UTR, and tri-nucleotide repeats being more evenly distributed along the gene (37%).

A 17,472 feature Actinidia oligonucleotide microarray was constructed based on sequence data from the Actinidia EST database. The 44–55 mer oligonucleotides were derived from gene sequences of different Actinidia species (Additional file 4). To test the cross hybridization between the different genomes of the Actinidia species, genomic DNA from two Actinidia species (A. deliciosa and A. eriantha) was hybridized to two different microarray slides. A plot of average signal intensity for the two species against each other is shown in Fig. 2. Of the 13,443 informative features, the majority (98.4%) showed a similar level of hybridization between the two species (less than an average of 2-fold change), with 172 showing a higher signal intensity in A. deliciosa (plus symbols, Fig. 2), and 44 showing a higher intensity in A. eriantha (triangles, Fig. 2). There was no correlation between the species from which the oligonucleotide was derived and the increased or decreased level of hybridization (Table 2). Consequently, the increases are probably due to separate gene duplications in each species.

The TC with the greatest number of ESTs was a cysteine proteinase (actinidin family, 758 members) followed by a metallothionein (547 members) (Additional file 5). However, when similar proteins were taken into account (i.e. different TCs that all matched a common Arabidopsis protein as their closest homolog with E < 1.0e^-100), there were 1,266 ESTs matching cysteine proteinases followed by 769 metallothionein protein-encoding ESTs.

It was not surprising to identify the cysteine protease actinidin as the most highly abundant TC cluster in Actinidia. Actinidin can constitute up to 50% of soluble protein in mature A. deliciosa fruit at harvest 21. Recent research has shown that this cysteine protease exists in basic and acidic forms and the amount of each form varies between species, with fruit of the important A. chinensis cultivar 'Hort16A' being almost devoid of acidic actinidin 22. ESTs for actinidin occur throughout the fruit libraries that make up the database, and these cluster into 10 distinct genes/alleles 22. As two of the key fruit libraries were subtracted for actinidin, the number of ESTs in the TC is likely to be an under-representation of the actual number. The function of this protease may be related to insect defense 23.

The metallothionein TC cluster contains a gene (pKIWI504) previously identified in kiwifruit 24 as being highly expressed in young fruit with reduced expression in the later stages of fruit development. However, the in vivo function of the metallothionein TC cluster is currently unknown. Interestingly, pineapple also had a high proportion of metallothionein ESTs 11.

Fourteen TCs in Actinidia with their proteins annotated by BiolView as "not assigned-unknown" proteins (i.e. matches to sequences with unknown function from Arabidopsis or other species) were highly expressed with over 100 EST members. Four of the unassigned TCs contained more than 300 EST members. These four TCs included kiwellin 25, which is a homolog to a grape ripening-related protein 26 with no Arabidopsis homologs (527 EST members); a homolog of At5g11420, a protein of unknown function (339 EST members); a senescence-related protein (no Arabidopsis matches, 311 EST members); and an ABA stress-related protein (no Arabidopsis matches, 302 EST members).

Functional analysis of ESTs and NRs using Mapman 27 is shown in Table 3. The analysis was conducted by identifying each NR's nearest homolog to Arabidopsis, using BLASTx and identifying domains/families that are indicative for a given function using InterProScan. Then that NR and its member ESTs were assigned to the corresponding bin of the Arabidopsis match. Of the 41,858 NRs, 28,345 had sufficient homology to an Arabidopsis sequence to assign it a functional classification in Mapman. Thus, 32% of NRs from Actinidia had no Arabidopsis homolog, but a proportion would have homologs in other crops (e.g. genes related to fruit ripening).

A comparison of the percentage of NRs assigned to the top level Mapman classification bins and the percentage of Arabidopsis genes classified in the same bin showed a strong relationship with an r² value of 0.96 and a slope of 0.87, near to the expected 1:1 slope. The same relationship when calculated for second level bins gave an r² of 0.98 and a slope of 1.14 (second level bin data not shown). Taking only those top level bins that contained more than 100 NRs to remove chance as a major factor, only two bins appeared to be over-represented (i.e. > 2.0 times higher than the mean) in Actinidia, namely bin 4 (glycolysis) and bin 8 (TCA cycle). This may reflect a high level of primary metabolism in fruit. Interestingly, when all NRs are included, bin 5 (fermentation) and bin 6 (gluconeogenesis) were added to the list of over-represented NRs. The same relationships for potato showed no anomalies (Table 3). This analysis demonstrates that the clustering of ESTs generated a fair representation of the distribution of genes.

Overall, the average number of ESTs per NR in the database was 4.0. There were only two bins with a high number of ESTs per NR, bin 15 (metal handling) with over 17 ESTs per NR and bin 22 (polyamine metabolism) with over 9. However, bin 22 had a low number of NR and EST members, suggesting this result may have occurred by chance. In contrast, bin 15 included the metallothioneins and the very highly expressed EST discussed earlier (Additional file 5).

TCs with high numbers of ESTs (> 100) came from 18 of the 35 Mapman bins 27 with approximately three quarters falling into the bin names "not assigned, no ontology", "protein degradation", "metal handling binding chelation and storage", "stress biotic", "cell wall degradation" and "secondary metabolism isoprenoids".

The most common InterPro families 28 were also analyzed (Additional file 6). There were 3,111 InterPro families (flagged as true by InterPro) represented in the Actinidia NRs, with the most NRs found in the protein kinases grouping. Following that were thioredoxins, while cysteine proteinases, which had the highest number of ESTs in the database, were represented as the 7^th most frequently occurring NR.

The most abundant volatile compounds in the flowers of these Actinidia species were alcohols and terpenes. A. chinensis flowers contained 71% sesquiterpenes, largely dominated by farnesols (Additional file 7). A. deliciosa flowers had the highest volatile ester content at around 12%, and aromatic (6–9%) and straight-chain alcohols (9%). These flowers were notable for their α-farnesene (12–26%) and germacrene D (0.3–12%) content. A. arguta flowers contained branched-chain (13%) and aromatic alcohols (14–50%), and monoterpenes (12–24%), which were mostly linalool oxidation products reported previously 29. A. eriantha flowers contained few volatile compounds, the notable ones being 2-phenylethanol (8.5%), 6-methyl-5-hepten-2-one (also 8.5%), and β-myrcene (5%). The flowers from two further Actinidia species showed distinct volatile profiles. A. polygama flowers contained linalool oxides (36.5%) and dehydroiridodial (37%), a monoterpene related to the Nepeta cataria (catnip) compounds 3031. The flowers of A. indochinensis flowers were dominated by 2-phenylethanol (25.8%) and a couple of terpenes, the major one being linalool (23.6%). The aromas from these flowers have also had descriptors attached (McNeilage et al. unpublished observations). For example A. eriantha has been described as coconut and berry, and A. arguta with descriptors including sweet, lime, rose and vanilla 29. A chinensis and A. deliciosa are described as having tea-rose aromas, sweet, fresh, lilac and violet, while A. polygama has gardenia and lime descriptors. These descriptors reflect the diversity of volatile compounds found in the Actinidia flowers.

In fruit, esters accounted for up to 85% of the total volatiles. The main A. chinensis esters were the fresh, fruity ethyl butanoate which, depending upon analytical method and genotype, ranged from 3 to 54%, and butyl butanoate, both of which are common fruit esters 32. A. chinensis 'Hort16A' fruit contained significant amounts of eucalyptol (0.2 to 21%) which has been described as having a fresh, pungent cooling taste 32. Acetaldehyde, hexanal, E-2-hexenal and ethyl butanoate are known to be important contributors to Hort16A aromas 33. Ethyl butanoate along with methyl butanoate and ethyl acetate were the major (~60%) compounds in the green A. deliciosa fruit and are known to be major volatile contributors to its flavor 34. The other major compound was ethanol (16%), with only trace levels of other alcohols and terpenes. Whilst at more modest levels in A. arguta fruit (2–8%), ethyl butanoate was still substantial. A. arguta fruit were notable for their content of methyl- and ethyl benzoate, which are responsible for the distinctive aroma of feijoa (Feijoa sellowiana O. Berg) fruit 35, and also camphor (7%). The dominant ester in A. eriantha fruit was butyl acetate (12%), a flavor compound found in many fruit, which has a strong, fruity odor and a taste reminiscent of pineapple 3236. However, the total volatile level in A. eriantha is lower than other species and the number of esters is also lower (Additional file 7).

Esters are synthesized from their acid and alcohol precursors by acyltransferases (ATs). We identified 30 ATs in the Actinidia EST database, of which 25 are full length. These sequences were analyzed for their phylogenetic relationship with other plant ATs of known function (Fig. 3). Two clades were identified that contained enzymes involved in the synthesis of flavor-related esters (Clades 1 and 2 in Fig. 3). The first contained alcohol acyltransferases (AATs) from apple, melon and banana and nine AATs from Actinidia (AdAT1, AdAT2, AeAT9, AcAT15, AcAT16, AdAT17, AaAT18, AcAT20, AcAT23), while the second clade contained AATs from strawberry and three AATs from Actinidia (AdAT6, AdAT22, AcAT24). An Actinidia AT predicted to be involved in esterifying anthocyanins (AcAT25) and two ATs predicted to be anthranilate acyltransferases (AdAT8, AcAT21) were also identified. Three Actinidia ATs were closely related to the carboxylesterase-related protein 2 CER2 and therefore predicted to be elongases (AcAT30, AcAT27 and AcAT26) possibly involved in wax production.

Multiple sequences of the same genes were recovered for some genes reflecting orthologs, alleles or sequences from different genomes within polyploid genomes. For example, the three ATs, AcAT16, AdAT17 and AdAT1, and the two ATs, AaAT18 and AdAT2, are probable variants of just two different AATs. A. deliciosa is a hexaploid (6n) possibly explaining this multiplicity of genes. We may have uncovered at least one instance where, for a single AAT, all three genomic versions have been isolated (AcAT16, AdAT17 and AdAT1).

Carboxylesterases conduct the opposite reaction to ATs, i.e. the hydrolysis of esters into acids and alcohols 37. As well as ester hydrolysis, plant carboxylesterases have been implicated in isoflavonoid biosynthesis 38, plant defense 39, and hormone regulation 40. Analysis of Actinidia flowers and fruits showed a wide range of acids and alcohols (Additional file 7).

Previously published plant carboxylesterase (CXE) genes from GenBank were used to identify CXE genes in the Actinidia EST database using BLAST searches. The sequence alignment of CXEs revealed amino acid motifs and secondary structural features characteristic for members of the α/β hydrolase fold superfamily and more particularly of CXEs such as an active site serine surrounded by a GXSXG motif.

In Actinidia 19 CXEs were identified, 18 of which are full length (Fig. 4). These include CXE members likely to encode gibberellin receptors (AdCXE9, AdCXE17, AdCXE15), an isoflavanone dehydratase (AeCXE11) and an ortholog of a plant defense-associated CXE (AcCXE12). Other CXEs from Actinidia align well with Arabidopsis CXEs and can be associated with the classes identified elsewhere 41. The structure of an Actinidia CXE is now available (AeCXE1), the first plant CXE to have its structure solved 42. AeCXE1 can hydrolyze a range of substrates from C2 to C16 esters, with a preference for C4 moieties. There are two CXEs from Actinidia (AdCXE16, AaCXE2) that may be orthologs to AeCXE1 from A. eriantha.

Actinidia terpenoid compounds such as monoterpenes (C10), sesquiterpenes (C15) diterpenes (C20), triterpenes (C30), carotenoids, sterols, phytols and quinones are derived from two common precursor molecules, isopentenyl diphosphate and dimethylallyl diphosphate. These precursors are produced via the cytoplasmic mevalonate pathway or the chloroplastic 1-deoxy-D-xylulose-5-phosphate (mevalonate-independent) pathway 43. Genes for enzymes in both these pathways are present in the Actinidia EST database, with the results for analysis of ESTs in the early steps of the mevalonate pathway shown in Fig. 5. The conversion of these precursors into other intermediates of the terpenoid biosynthetic pathway, including geranyl diphosphate, farnesyl diphosphate, and geranylgeranyl diphosphate, is carried out by polyisoprene synthase genes (EC 2.5.1.x). Over 200 ESTs with sequence homology to known polyisoprene synthase genes were found in the Actinidia EST database including 116 ESTs with similarity to dimethyallyltranstransferase (EC 2.5.1.1), 66 ESTs with homology to known monoterpene and sesquiterpene synthases and 38 ESTs homologous to squalene synthase (EC 2.2.1.21). No ESTs for phytoene synthase (EC 2.5.1.32) were identified.

The EST database collection has been used to identify a multifunctional terpene synthase gene from A. deliciosa flowers that produces the sesquiterpene germacrene D and a range of other sesquiterpene products at lower abundance 44. This gene is represented by 10 sequences in the EST database.

Green fleshed commercial kiwifruit such as A. deliciosa 'Hayward' are characterized by the retention of chlorophylls in the flesh of ripe fruit 45. In contrast, the newer yellow-fleshed fruit cultivars (e.g. A. chinensis 'Hort16A' and 'Jintao') degrade the chlorophyll during ripening and the consequent loss of the green color uncovers the underlying yellow carotenoid pigmentation.

The degradation of chlorophylls (leading to the production of colorless breakdown products) is controlled by catabolic enzymes in the chlorophyll degradation pathway 46. Actinidia ESTs discovered in the pathway of chlorophyll breakdown are shown in Fig. 6. Chlorophyllase is the first enzyme of the catabolic pathway and it removes the phytol chain from chlorophyll a. Homologs of this enzyme have been mainly identified in libraries of vegetative tissues (breaking buds, leaves and petals) of A. deliciosa and A. chinensis. Homologs of pheophorbide a oxygenase have also been identified, with eight ESTs representing two genes from fruit libraries of A. chinensis. This enzyme is responsible for opening the tetrapyrrolic ring of the chlorophyll molecules, the critical step for the loss of green color. Surprisingly, we did not identify any ESTs of red chlorophyll catabolite reductase, an enzyme associated with pheophorbide a oxygenase and required to complete the porphyrin ring opening and the degradation of chlorophyll.

The fruit of many Actinidia species contain a range of carotenoid pigments, including β-carotene and lutein 45. The yellow-fleshed kiwifruit A. chinensis 'Hort16A' develops as a green-fleshed fruit, but on ripening the chlorophyll is degraded exposing the carotenoids in the yellow flesh. ESTs for most steps in the biosynthesis of carotenoids 474849 are present in the Actinidia EST database (Fig. 7), except the first step, phytoene synthase. Ninety eight ESTs for phytoene desaturase were identified, with 94 coming from A. deliciosa libraries, and the remainder from A. chinensis libraries. Of these 98 ESTs, 90 came from a single A. deliciosa fruit library with eight from bud and meristem libraries. Of the 56 ESTs found for lycopene β-cyclase, 23 were from petal tissues and 32 from fruit libraries. These ESTs were largely found in libraries from A. deliciosa (31) and A. chinensis (24), with only one EST from A. eriantha. In contrast, only a single EST was found for lycopene ε-cyclase, which is the enzyme required with lycopene β-cyclase to convert lycopene to α-carotene 50. There were five ESTs of β-carotene hydroxylase found in petal (3), leaf (1) and dormant bud (1) libraries, but only a single EST was present for ε-carotene hydroxylase. Both of these enzymes are required for lutein synthesis 51.

Flavonoids are polyphenolic secondary metabolites synthesized through the phenylpropanoid pathway 52. Anthocyanins are a subgroup of the flavonoids, which appear red to blue, depending on the pH, due to additions to their phenolic rings. In red fruit of A. chinensis and A. deliciosa, the predominant anthocyanins are cyanidin-based and the preferred glycosylation is in the 3' position 53. While many of the kiwifruit libraries came from tissue with low levels of anthocyanins, several libraries were derived from tissue containing anthocyanins, including red fruits, brown skins and pink petals, leaves and buds (Additional file 2a).

The anthocyanin biosynthetic pathway was analyzed by BLAST searching for ESTs representing different enzymes in the pathway (Fig. 8). Chalcone synthase (CHS) was a highly represented gene family in the EST database with ~300 ESTs. There appeared to be three distinct CHS NRs, possibly representing three genes, that are found in mostly young A. eriantha fruit libraries, as well as in bud and petal libraries. There was only one apparent chalcone isomerase (CHI) NR and over half the ESTs for this gene were found in shoot and bud libraries. Two distinct flavanol 3-hydroxylase (F3H) NRs were found, with ESTs predominantly found in bud and cell culture libraries. Only one flavonoid 3',5'-hydroxylase (F3'5'H) NR was found, from A. eriantha fruit, while ESTs for two distinct flavonoid 3'-hydroxylase (F3'H) NRs were predominantly found in shoot and bud libraries. Of the four potential dihydroflavonol reductase (DFR) NRs, the most highly represented was found in bud and young fruit. Three NRs, perhaps representing three gene families, of leucoanthocyanidin dioxygenase (LDOX) were found in the EST database, with most ESTs in bud libraries. The ESTs with highest sequence similarity to UDP-glucose flavonoid 3-O-glucosyltransferase (UFGT) are also predominantly found in developing buds of A. deliciosa.

Many of the ESTs for anthocyanin biosynthesis were sequenced from young fruit libraries (particularly A. eriantha) and in A. deliciosa buds. Developing buds from A. deliciosa appear to be a darker red color and therefore presumably accumulate more anthocyanins than buds from A. chinensis and A. eriantha, so it is not surprising to find ESTs for these later biosynthetic steps in bud tissue from A. deliciosa. However, ESTs for genes encoding enzymes from the early steps in the flavonoid pathway (e.g. CHS and F3H) are found in unpigmented A. eriantha fruit suggesting these tissues may accumulate a different class of flavonoids.

Most of the enzymes for anthocyanin biosynthesis are well represented in the EST database except for F3'5'H. This is not surprising given the only anthocyanins identified in A. deliciosa and A. chinensis are glycosylated in the 3' position. There are, however, species within the genus Actinidia, such as A. melanandra, that accumulate high levels of delphinidin, a different aglycone that would require F3'5'H activity. No ESTs from these species were sequenced.

Kiwifruit are well known to be high in ascorbic acid (vitamin C), with values in commercial varieties ranging from 80 to 120 mg per 100 g fresh weight of fruit 54, and with A. eriantha having up to 10 times as much again. Ascorbic acid is synthesized in plants from glucose, through three potential pathways: the L-galactose pathway, the galacturonate pathway and through myo-inositol 55. Myo-inositol is a major sugar alcohol found in kiwifruit (Additional file 7; 5657). The only pathway that is well established is the L-galactose pathway for which all the enzymes have now been identified in plants 5558. Many of the genes are known for the other two potential pathways, but several steps are still unknown. The Actinidia EST database was searched for ESTs similar to genes known to encode steps in these pathways. The numbers of ESTs homologous to various identified genes in the three pathways of ascorbic acid biosynthesis are shown in Fig. 9. Surprisingly for these high vitamin C plants, while the number of ESTs found in early steps in all the pathways is relatively large, the number of ESTs found in the two last steps is much lower, with no examples of ESTs encoding the last step of two of the pathways, that catalyzed by L-galactono lactone dehydrogenase.

Fruit from Actinidia species have a relatively high total acid content (1–3% w/w) of which 40–60% can be quinic acid, 40–60% citric acid and 10% malic acid (Additional file 7). Quinic acid comprises an even higher proportion of the total acids during early fruit development 59, and over 25-fold variation in quinic acid content can be found in fruit of a single A. chinensis cross 60. Kiwifruit are unusual in having such a high content of quinic acid as a fresh fruit, although levels of quinic acid in cranberry juice can be around 1% 61 and peaches can have up to 0.2% quinic acid 62.

Quinic acid is a key intermediate in lignin biosynthesis, folic acid metabolism, aromatic acid synthesis, anthranilate biosynthesis, and purine metabolism. The molecular and enzymatic control of quinic acid storage and metabolism may affect all these pathways. However, little is known about quinic acid metabolism in Actinidia species. ESTs for all the enzymes in the quinic acid biosynthetic pathway to shikimate (Fig. 10), except for a quinate dehydrogenase, have been identified in the Actinidia EST database.

Of the 13 ESTs identified for dehydroquinate synthase, eight were found in fruit or petal libraries, the rest were sequenced from bud or leaf libraries. Of the 16 ESTs identified for the bifunctional dehydroquinate dehydratase/shikimate dehydrogenase, 11 were from fruit or petal libraries and five from bud or leaf libraries. These ESTs were found mainly in libraries made from A. deliciosa and A. chinensis tissues and to a lesser extent in A. arguta and A. eriantha libraries, reflecting the distribution of ESTs among the species. An enzyme has been reported with predominantly quinate dehydrogenase activity 63, but the gene for this enzyme has not been cloned.

Food allergies are estimated to affect ~6% of young children and ~3% of adults 64. Kiwifruit have been recognized as a potentially allergenic fruit for over 20 years and reported allergies to kiwifruit are increasing 65. Most allergy symptoms to kiwifruit are quite mild, but severe reactions have been reported, particularly in young children 66. Allergies have been recorded in all three commercial species of Actinidia 67.

The Actinidia EST database contains sequences with homology to many known plant allergen proteins 68. Potential allergens include the 2S albumin proteins, lipid transfer proteins, thaumatin-like protein (TLP), α-amylase/trypsin inhibitors, latex allergens, plant chitinases, profilins, cystatins, Bet v 1 homologous proteins and plant seed globulin allergens. However, only a small number of potential kiwifruit allergens have been directly confirmed by immunological testing, including the cysteine protease actinidin (Act d 1; 69), an unidentified 43 kDa protein (Act d 2; 70, a TLP 70 and kiwellin 25.

ESTs for actinidin occurred throughout the 37 libraries from seven species that make up the database, and these clustered into acidic and basic forms of 10 distinct genes/alleles 22. No matches were found by BLAST searching the EST database for the 43 kDa protein Act d 2. Kiwellin ESTs were found in libraries from most Actinidia species but were particularly abundant in A. eriantha ripe fruit skin, accounting for ~40% of ESTs in that library. The EST database contained 24 NRs with homology to TLPs from other plants (e.g. Pru a 2 from cherry and thaumatin 1 from Thaumatococcus daniellii), and all but one EST contained the 16 conserved cysteine residues characteristic of this protein class 71. The various NRs could be aligned and clustered into 4 or 5 groups (data not shown), with two NRs accounting for the bulk of the ESTs (46 and 29%).

Globulin proteins (11S and 7S) have been reported from a wide range of seeds, including cereals 7273 and legumes 74. Globulin-like proteins are also found in A. thaliana 75 and a gene family of 10 globulin-like genes are present in Arabidopsis gene databases (The Arabidopsis Information Resource (TAIR), http://arabidopsis.org/). 11S globulins are composed of sets of polymorphic subunits derived from a multigene family. They exist as hexamers with molecular mass between 300 and 400 kDa, or as trimers. 11S globulin subunits consist of two polypeptide chains linked by at least one S-S bridge between cysteine residues at highly conserved positions in the acidic alpha-chain and basic beta-chain. Both chains are post-translationally generated from a common precursor protein, which represents the product of one member of the multigene family 76. 11S globulins have been shown to be related to the 7S globulins through sequence comparisons 77.

A BLAST search of the Actinidia EST sequence database with a range of previously identified globulin protein sequences identified two ESTs with a predicted protein sequence matching 11S globulin, both from a library made from A. chinensis whole fruit, which included seeds. Other related sequences from Actinidia fruit libraries belonged to the 7S vicilins-like family (38 ESTs), which were found in only fruit-based libraries from A. deliciosa, A. chinensis and A. arguta. These libraries would have all included seeds. A third grouping of sequences was identified that we have named other globulins (OG) (201 ESTs). The OG group appears to be more closely related to the 7S globulins than to the 11S (Fig 11). All three groupings are supported by the clustering of the Arabidopsis globulin genes as well as globulin genes from other species. While most OG ESTs were found in fruit libraries (81%), which included seeds, some members of the OG family were found in fruit skin, buds and petal libraries, suggesting a role other than as seed storage proteins. These OG ESTs were found in libraries derived from A. deliciosa (100), A. chinensis (42), A. arguta (39), and A. eriantha (16).

Both headspace and solvent sampling was used to obtain volatiles. For headspace sampling of flowers, whole flowers (2.5–5 g) that were 50–75% open were placed in a 250 ml Quickfit™ Erlenmeyer flask to which was fitted a volatile trapping cartridge packed with 100 mg of Chromosorb 105™ absorbent and sampled and analyzed as described earlier 29. For fruit, ten fruit (eating-ripe) were cut longitudinally in half and one half of each fruit was used for headspace sampling and analysis 29. The other halves were frozen in liquid N₂ for solvent extraction.

For solvent extraction, either ~1 g of flower petals were rinsed in 2 ml of purified 50:50 pentane/Et₂O or 20 g of fruit was pulped, and gently shaken in a sealed test tube with 10 mL of pentane:Et₂O (8:2) for 1–2 min, four times over a 2 h period at room temperature. Samples were processed and analyzed by GC-MS as described earlier 29.

Quantification of compounds was carried out using an average FID response based on methyl butanoate, ethyl butanoate, hexanol and methyl benzoate. Component identification was by comparison with spectra in the Mass Spectral Database (1998 NIST and an in-house database), retention indices (in-house database) and in some cases direct GC-MS comparison with authentic standards.

Tissue was ground in liquid N₂ and acids and sugars were extracted from a known mass of tissue (1.4–2.5 g) into 10 mL of 80% EtOH (with adonitol and tartaric acid added as internal standards) at 60°C for 1 h. The extracts were processed, sugars and organic acids derivatized and then quantified by GC 97. Identification of the acids and sugars was confirmed by GC-MS comparison with mixtures of authentic compounds.

Sequence data for the ESTs described in this paper can be found at GenBank under accession numbers FG396013 – FG528589.