Transcriptome and proteome
Large-scale analyses of gene expression can provide information on the regulation of gene networks underlying metabolic and cellular processes. Transcript abundance primarily reflects the transcriptional control of gene expression and has been shown in many instances to be poorly related to protein abundance (Gygi et al., 1999; Watson et al., 2003). Proteins are the actual cellular effectors. Their abundance depends on the balance between translation and degradation which, in turn, can be influenced by various co- and post-translational processes. Comparison of the distribution of transcripts and proteins in the different functional categories at a given developmental stage can shed light on the relative importance of these processes. Integration of the temporal dimension may help to reveal sets of co-regulated gene products and to decipher the respective roles of transcriptional and (co- and post-) translational control on elements of the networks.
To date, few investigations have combined transcriptomic and proteomic data. The most recent reports have focused on small numbers of proteins in special situations, such as seed protein processing (Higashi et al., 2006) or microspore-derived embryo development (Joosen et al., 2007). The transcript analysis in this article deals with several hundred clones from cDNA libraries at three critical stages of endosperm development. As expected, the major changes in the transcriptome between the three stages involve the increased proportion of storage protein transcripts, but also the decrease in NYC and NI transcripts, suggesting that a large number of endosperm-specific functions are yet to be deciphered at the early stage. A proteome map consisting of 632 proteins has been established previously for the 14-DAP endosperm of the F-2 genotype in the same EU–Zeastar programme as the transcriptome analyses (Méchin et al., 2004), with the protein identifications updated recently (Méchin et al., 2007). The distribution of the functional categories – excluding storage proteins and NYC and NI transcripts – of the 10-DAP transcriptome is closer to that of the 14-DAP proteome, revealing a lag phase between gene transcription and protein accumulation. The proportion of sequences in the metabolism category is higher in the proteome data than in either transcriptome data set (40% vs. 21%–28%). This observation, associated with the high proportion of isoforms for a given function generally observed in proteome data, suggests that a number of post-translational modifications are possible for these accumulating proteins.
The combined protein synthesis and RNA processing category is a major constituent of the transcripts throughout the time course. RNA processing proteins are absent in the proteome analysed at 14 DAP, probably indicating high turnover with low accumulation. The RNA processing transcript proportion increases at 14 and 21 DAP, whereas, at 10 DAP, the high proportion of protein synthesis-related transcripts (87 of 105) appears to precede the accumulation of proteins of this category at the early to mid developmental stages (Méchin et al., 2007). These observations may be related to the increasing translation and protein synthesis activity needed to support zein production, beginning at about 14 DAP (Méchin et al., 2007). The protein destination category is also a major category over the time course, with the proportion of transcripts devoted to degradation increasing up to 14 DAP and decreasing afterwards (44%, 53% and 33%, respectively). This trend appears to be consistent with the maximum accumulation of proteins involved in degradation processes at the early stages (Méchin et al., 2007). The relatively high proportion of expressed genes devoted to degradation over the time course is indicative of important protein turnover during endosperm development, with a maximum at 14 DAP. Although it is not surprising to see this transition from the lag phase to the accumulation phase when the set of enzymes must be renewed to enable new functions, it is less expected during the phase in which the starch accumulation rate is constant. Of the transcripts coding for protein-folding genes, another subcategory of protein destination, it is tempting to associate the progressive and large accumulation of one specific protein, disulphide isomerase (gene index identification for protein: 1709619), up to 30 DAP with folding of the most abundant protein, i.e. zein. Transcription-related transcripts are the most abundant at 21 DAP, which may be related to the preparation of the maturation phase, which progresses after 30 DAP. Transcripts for defence (antifungal and antimicrobial proteins, etc.) and detoxification enzymes against reactive oxygen species (ROS) and xenobiotics are always present, and the abundance of proteins identified in this category at 14 DAP reaches a maximum during the earliest stages and diminishes afterwards. The presence of an anti-ROS protein at this early stage can be explained by the large oxygen availability, as opposed to the hypoxic situation which prevails after 14 DAP (Rolletschek et al., 2005; Méchin et al., 2007). However, superoxide dismutase (SOD) transcripts have been reported at 28 DAP, and their expression is enhanced by xenobiotics (Mylona et al., 2007). It has been suggested that this expression may be related to seed maturation. This would be consistent with the presence of metallothioneins and glutathione S-transferase transcripts at 21 DAP, which are also associated with this process (Table S1, see ‘Supporting information’). The communication/signal transduction functional category is represented more by transcripts (12%–15%) than by proteins (approximately 1%) at 14 DAP. A probable explanation is that these transcripts, which mostly code for kinases and transduction chain elements, are not detected on two-dimensional gels because of their low abundance.
We investigated the metabolism category in more detail in order to examine the relationships between energy metabolism and storage compound synthesis. The decrease in TCA cycle and mitochondrial transcripts to their disappearance at 21 DAP, and the appearance of the PPDK transcript, are consistent with the observations at the protein level, although the variation is more dramatic and contrasting for transcripts than for proteins. These data can be interpreted by the near absence of oxygen at these late stages, which impedes functioning of the mitochondrial electron transport chain because of the deficit in the final electron acceptor. In this context, starch synthesis, which does not need oxygen, is favoured. In addition, the increase in PPDK may act on the protein/starch balance at the end of starch filling, as discussed by Méchin et al. (2007). The largest diversity of starch and amino acid metabolism transcripts is observed at the beginning of the starch accumulation phase, preparing for the synthesis of the numerous enzymes needed for further storage product synthesis. The large increase in carbohydrate transcripts at 21 DAP corresponds to UDP-glucose metabolism enzymes (conversion into UDP-galactose or UDP-glucuronate) and to glucosidase, which is probably linked to cell wall formation. Accordingly, a relatively high UDP-glucose level is maintained at later stages. With regard to the proteome, especially at the mid stage, the proportion of transcripts for secondary compounds and nucleotide metabolism is surprisingly high compared with those for carbohydrate metabolism, suggesting that the contribution of these two categories of metabolism may need to be re-evaluated.
Starch synthesis: towards a synthetic view from transcripts, proteins and enzyme activities
The core reactions involved in the starch synthetic pathway are well known, especially in maize kernels, thanks to a series of mutants marking each step. Individual analyses of transcript and protein expression have been made by classical techniques, and post-translational modifications, possibly affecting in vivo activity, have been demonstrated for some enzymes, such as redox modification in AGPase and phosphorylation in SBE (Tetlow et al., 2004). The coordination of individual reactions has been tested genetically by assessing the effect of single mutations on the rest of the pathway. Taking advantage of the triploid structure of the endosperm, the dosage effect of mutant alleles of Sh1, Bt2, Ae1 (encoding SBEIIb) and Su1 (encoding DBE) on the content of soluble sugars and starch formation has been examined for various enzyme activities (Singletary et al., 1997). These workers found a direct relationship between the amount of starch produced and gene dosage, and observed, for all three dose mutants, except su1, large pleiotropic effects on the various enzyme activities, associated with an elevated level of soluble sugars. Similarly, sh2, bt2, ae1, du1 and waxy mutations resulted in both an increased transcription rate of the non-mutated genes of the starch pathway and an effect on zein synthesis, which suggested a coordinated transcriptional control of the genes involved in storage compound biosynthesis in maize endosperm (Giroux et al., 1994). In rice, the study of 269 genes belonging to different metabolic pathways that were coordinately up-regulated during grain filling, as shown by the monitoring of 21 000 microarrayed genes, has recently provided some clues on the possible mechanisms involved in coordinated expression (Zhu et al., 2003). Indeed, analysis of available promoter sequences indicated a significant over-representation of an AACA element in these grain-filling genes when compared with other genes, suggesting that this motif may play a role in the coordinated expression of various pathways during grain filling. A group of transcription factors potentially interacting with this element was identified.
The present approach in maize, which considers the simultaneous developmental changes at key steps of starch biosynthesis at the level of transcript and protein accumulation and enzyme activity, in relation to substrate and/or product concentration, provides information on the integrated functioning of the pathway during endosperm development and its relationship to the energy metabolism examined at the transcript and protein level.
Two main patterns are observed for carbohydrates: (i) hexoses present an early maximum followed by a decline; and (ii) starch, ADP-glucose and sucrose present a low level before 14 DAP and a progressive increase up to 30 or 40 DAP. The enzymes involved in the cleavage/synthesis or accumulation of these compounds display consistent activity patterns. Initially high invertase activities decline thereafter, whereas the activities of the core enzymes of the starch pathway (Susy, AGPase, GBSS, SSS and SBE) progressively increase from a low initial level to a maximum at 30–35 DAP, in parallel with the ADP-glucose content. These two distinct patterns may be interpreted with regard to the contribution of the metabolites to development, either as signalling molecules or as trophic components, during the lag and accumulation phases.
Invertases play a prominent role during the early phase, producing hexoses from imported sucrose, as inferred from the analysis of the miniature phenotype produced by the mn1 mutation. This mutant is characterized by a deficiency in both soluble and cell wall invertase activities, although genetic and molecular analyses show that the mutation only affects the cell wall gene Incw2. This gene is specifically expressed in the basal endosperm tissue cell layer, where carbohydrate transfer from the maternal placento-chalazal zone to seminal tissue takes place (Cheng et al., 1996). As cell wall and vacuolar activities vary in parallel in the mn1 mutant, a coordinated control of both isoforms has been suggested. We confirm that the cell wall form is the most active invertase in the endosperm, as noted previously (Cheng et al., 1996; Qin et al., 2004). Four genes encode cell wall invertase (Taliercio et al., 1999; Kim et al., 2000), but only Incw2 and, to a much lesser extent, Incw1 are expressed in the kernel. Accordingly, the main transcript detected is Incw2, with maximum expression appearing earlier than activity. However, the activity is relatively high at 10 DAP, but the transcript level is still low, suggesting that, at the early stage, few transcripts are required, and that, afterwards, the cell wall invertase inhibitor recently reported in maize kernels hinders activity (Bate et al., 2004). Invertase activity appears to be critical for early kernel development (Cheng and Chourey, 1999) because hexoses act as a positive signal on cell division (Vilhar et al., 2002) and α-tubulin transcript expression (Datta and Chourey, 2001). Accordingly, α-tubulin transcripts are most numerous at 14 DAP and proteins are mostly expressed in the lag phase (Méchin et al., 2007).
Susy and AGPase are two multimeric enzymes encoded by small multigene families with specific tissue expression. Sh1 for Susy and Bt2 and Sh2 for AGPase are the main genes expressed in the endosperm. Transcripts for two other Susy genes, Sus1 and Sus3 (now called Sus2), have been reported in the endosperm (Carlson et al., 2002). Indeed, very low expression of these two genes compared with Sh1 was found, but quantitative proteome data were only available for Sus1, showing maximum expression at 30 DAP, consistent with the enzyme activity pattern. The major transcripts of Sh1 peak at 21 DAP, which is earlier than for the enzyme activity, consistent with a predominant role for transcriptional control of the expression of Susy in the endosperm, as also reported in developing leaves (Nguyen-Quoc et al., 1990; Nguyen-Quoc, 1991).
For AGPase, the situation is even more complex as, in addition to the small and large subunits encoded by the Bt2 and Sh2 genes expressed in the endosperm cytosol, another subunit couple, Agp1 and Agp2, constitutes an AGPase expressed in both the endosperm and embryo amyloplasts (Denyer et al., 1996), but mainly in the embryo (Giroux and Hannah, 1994). Transcripts for each AGPase subunit were identified for the embryo forms, but no endosperm Bt2 transcript was detected in our cDNA library. The RT-PCR results show that Sh2 is the most highly expressed overall. A similar predominant expression of only one AGPase transcript (OsAPL3, a Sh2-like gene) in the middle phase of seed development has been reported from comparisons of the expression profiles of six cDNAs encoding the two small subunits and four large subunits in rice. Furthermore, OsAPL3 has been shown to be accumulated significantly in response to increasing levels of sucrose and abscisic acid (Akihiro et al., 2005). The present sucrose time course is consistent with a similar control of Sh2 expression in maize endosperm. Classical Northern blot analyses with Bt2 and Sh2 probes of the time course have shown previously that relative variations of both transcripts coincide, but slightly precede AGPase protein and activity levels (Prioul et al., 1994), a result consistent with the present observations. It has also been noted, as in this study, that the maximum activity coincides with the maximum rate of starch accumulation. The low level of the AGPase protein spot assigned to the large subunit of the embryo (plastid) form at 10 DAP, in the absence of any measurable activity, may originate from a contamination of the endosperm by maternal tissues, which are difficult to separate at this stage. Accordingly, up to 14 DAP, the AGPase transcripts and peptides detected at low levels are mostly located in the pericarp and nucellus (Brangeon et al., 1997). Afterwards, the contribution of the plastid form to the total AGPase in the endosperm has been estimated to be approximately 10% (Denyer et al., 1996). In any case, the major form becomes the typical endosperm AGPase from 14 DAP onwards. Our in situ measurements of AGPase activity show that it is concentrated in the endosperm, first in the distal part and then towards the base. At 40 DAP, starch synthesis is completed in the upper endosperm, as shown by the yellow zone marking flint starch where the enzyme is no longer active. The pattern is very similar for Susy, as is the time course for both enzyme transcripts, supporting the cooperative functioning of both enzymes and suggesting a similar transcriptional control.
Three groups of enzymes are involved in starch synthesis [the starch synthases (GBSS and SSS), SBEs and, finally, DBEs hydrolysing branched linkages], yielding the highly ordered and clustered structure of the amylopectin polymer. Pullulanase, another DBE group, is more likely to be involved in starch degradation. Of the starch synthases, the granule-bound form (GBSSI) is dedicated to the synthesis of amylose, which consists of long linear chains with a few branches. The very low transcript and protein levels at 10 and 14 DAP increase further as amylose accumulation increases, which may be indicative of a coarse transcriptional control, although some post-translational modification cannot be excluded because of some discrepancies between the amlyose accumulation rate and both transcript and protein levels. The SSSs are involved in the elongation of amylopectin chains. Four SSS isoforms (I, IIa, IIb, III = DU1) have been described, but most of the activity has been reported to rely on SSSI and SSSIII products (Cao et al., 1999). However, of the three transcripts characterized here, SSSIIa is the most expressed overall, which is consistent with our proteomic data. Thus, the contribution of SSSIIa to global activity in the endosperm needs to be re-evaluated. Three SBEs (I, IIa and IIb) have been described (Ball and Morell, 2003); the transcripts assigned to SBEII are expressed three times more highly than SBEI, whereas SBEIIa is only weakly expressed, mainly in the early phase. The activity and transcripts nearly disappear at 40 DAP, suggesting rapid inactivation or degradation of the enzyme.
The present functional approach strongly suggests that the primary control of enzyme activity leading to starch biosynthesis takes place in a coordinated manner at the transcriptional level. The explanation may be found by searching for common specific motifs in gene promoters when complete sequences become available, as performed in rice (Zhu et al., 2003). However, in the case of cell wall invertase, AGPase and SSS for example, the joint analysis of the transcript, protein (when available) and activity level suggests a potential control at the protein level. The various controls [post-transcriptional modifications or feedback loops, hexose signalling, as shown for Incw2 and kernel cell division (Cheng and Chourey, 1999; Vilhar et al., 2002), and protein–protein interactions producing metabolic channelling (Hennen-Bierwagen et al., 2008)] remain to be elucidated.