Introduction

Post-harvest physiological deterioration (PPD) is an endogenous root disorder affecting the storage roots of cassava (Manihot esculenta Crantz). Within 24–72 h of harvest, depending on the cultivar and environmental conditions, roots exhibit a blue–black discoloration of the xylem vessels known as “vascular streaking” and display other unpleasant organoleptic properties such as impaired flavour and odour. This rapid deterioration is a major constraint on the development of this important crop, which impacts on farmers, processors and consumers alike.

Vascular streaking initiates at wound sites caused by harvesting (Booth 1976) and is thought to result largely from peroxidase mediated oxidation of scopoletin (Tanaka et al. 1983; Wheatley and Schwabe 1985). Biochemical localisation data indicate that required components of the reaction, namely scopoletin, H2O2, and peroxidase activities, are initially localised to the vicinity of the xylem vessels where vascular streaking symptoms occur. Moreover, peroxidase isoforms with activity towards scopoletin are present in the root (Reilly et al. 2003). Other observed biochemical changes during PPD include increases in respiration (Hirose 1986), changes in lipid composition (Lalaguna and Agudo 1989), ethylene synthesis (Hirose et al. 1984), the occurrence of a wound-induced oxidative burst (Reilly et al. 2003), and the accumulation of secondary metabolites (Rickard 1981; Sakai and Nakagawa 1988; Uritani et al. 1983; Wheatley and Schwabe 1985). Secondary metabolites which accumulate during PPD include diterpenic and phenolic compounds, several of which show anti-microbial and/or anti-oxidant activity (Buschmann et al. 2000a, b; Sakai and Nakagawa 1988).

At the molecular level the development of PPD is evidently a complex phenomenon involving multiple components. Cycloheximide inhibition of protein synthesis indicates PPD is an active rather than degenerative process involving changes in gene expression and protein synthesis. Several analyses of genes expressed during PPD (Han et al. 2001; Huang et al. 2001; Reilly et al. 2001, 2003) show altered regulation of proteins and enzymes involved in signal transduction, reactive oxygen species (ROS) modulation, phytohormone synthesis, senescence and programmed cell death (PCD) responses, synthesis of anti-microbial, anti-oxidant or other defensive compounds, and the formation of other compounds that are involved in the synthesis of cell wall components, as well as proteins that are themselves components of cell walls.

Under normal storage conditions wound healing of the root is inadequate; however, harvested roots are capable of producing a wound periderm under conditions of high temperature and humidity (curing), although this occurs slower than in other root crops. It has been proposed that wound repair and subsequent down-modulation of the wound signals are impaired, resulting in cascades of wound responses that spread throughout the root (Han et al. 2001). Interestingly, wounded roots that remain attached to the plant are capable of normal wound repair (Mwenje et al. 1998). Signalling components that could lead to the modulation of PPD responses include the plant hormone ethylene (Hirose 1986) and a wound-induced oxidative burst (Reilly et al. 2003). The role of hydrogen peroxide as a signalling molecule in plants is becoming increasingly clear and hydrogen peroxide may function to regulate several signalling pathways. Microarray data have indicated that around 1% of the Arabidopsis thaliana transcriptome is regulated by hydrogen peroxide (Desikan et al. 2001). This approach highlights the usefulness of exploiting massively parallel methods for gene discovery and analysis. Using a microarray of 5,700 cassava expressed sequence tags (ESTs) enriched for cDNAs expressed during starch biosynthesis and bacterial blight infection this approach has been successfully exploited to study the gene expression profile in cassava challenged with Xanthomonas axonopodis (Lopez et al. 2005). An analysis of the cassava proteome has also been used to explore protein accumulation during the developing storage root (Sheffield et al. 2006). In this paper we employ a cDNA-microarray constructed from cassava roots over a deterioration time course as a tool for gene discovery in order to identify PPD-responsive genes in the cassava storage root.

Materials and methods

Plant material

Cassava plants were grown in the tropical greenhouse at the University of Bath at 22–28°C, relative humidity 40–80%, and a minimum light period of 12 h per day under daylight, supplemented with 400 W Phillips high-pressure sodium lights when necessary. Storage roots (cultivar CM2177-2) were harvested after 9 months growth. Alternatively, roots were obtained from field-grown plants at CIAT, Cali, Colombia. Cassava leaf tissue was obtained from greenhouse-grown plants at the University of Bath.

Plant treatments

Induction of PPD in cassava storage roots was carried out as described previously (Reilly et al. 2003). For the leaf-wounding treatments, leaves were injured by crushing and were either immediately detached from the plant and stored at ambient temperature, or were left attached to the plant to allow initiation of wound repair. Samples were frozen in liquid nitrogen after 24 h.

cDNA library construction

Total RNA was extracted and mRNA purified from a range of time-points over a deterioration time course from roots of cassava cultivar CM2177-2 according to previously described methods (Reilly et al. 2001). Purified mRNA from time-points 0, 6 and 12 h after harvest were pooled as was mRNA from 24, 48 and 96 h and used to generate “early” and “late” PPD cDNA libraries in Lambda Zap II (Stratagene). These libraries should contain cDNA copies of all the genes expressed in the mature root at harvesting and during the first 4 days post-harvest.

Preparation and hybridisation of cDNA microarrays

Individual clones from the two PPD-related cDNA libraries were amplified by PCR and spotted on to PL-100C-Poly L lysine slides (CEL Associates) using a Hitachi SPBIO robotic arrayer. For each hybridisation 11,136 randomly chosen cassava cDNAs were spotted, 7,680 from the “early” and 3,456 from the “late” libraries. In addition, the following control DNAs were included on each slide: E1 (GenBank accession U89872); E2 (AF878810); E3 (NM_000858); E5 (K01193); E6 (X65316); E7 (AF126210); E8 (X13988); E9 (M21812); E10 (X07868), as “spike” genes; three well characterised genes from other plants (M76402, BE341179 and EF044211; and ten cassava genes with known PPD-expression profiles (EF035080, AF170272 AF239615, AY973618, EF035081, AF265551, AF266465, AF323605, AY036011 and AY033386). On each slide, every cDNA was spotted four times in order to generate four technical replicates per slide. As targets, mRNA was purified from roots of cultivar CM2177-2 over a deterioration time course, or from injured or uninjured leaves of the same cultivar. This mRNA was used to generate cDNA hybridisation targets labelled with either Cy3 or Cy5; cDNA synthesis was as Lopez et al. (2005). In order to obtain statistically reliable data, three biological replicate experiments were carried out for each hybridisation using independently isolated mRNA samples.

Microarray analysis

The PPD time course microarray data were analysed using the TIGR spotfinder and MIDAS software packages within the TM4 open source suite of programs (Saeed et al. 2003, www.tigr.org/softlab) using Lowess (Locfit) normalisation parameters applied globally for each slide followed by standard deviation regularisation between subgrids and slides. The mean (μ) and standard deviation (σ) for the distribution of log2 (ratio) values were calculated and differential expression at the 95% confidence level was identified as log2 (ratio) values greater than 1.96 standard deviations from the mean. For significance of microarray analysis of the microarray data of the comparisons between leaves and roots, gene expression was normalised using MIDAS (Saeed et al. 2003) by applying Lowess intensity-dependent normalisation by subgrids, followed by a standard deviation regularisation between subgrids and between slides. Genes expressed in the roots and not in the leaves at each root vs. leaf comparison were identified using the one-class design of the significance analysis of microarrays method (Chu et al. 2002), selecting 5,000 permutations and a false discovery rate of 0.1% or lower. The root-expressed genes common to the three comparisons were classified as root-specific.

RNA blot analysis

Total RNA for Northern hybridisations was extracted from storage roots or leaves and blotted and hybridised as Reilly et al. (2001). Hybridisation was carried out with 32P-labelled cDNA probes prepared using the Ambion Stripeze labelling kit according to the specifications of the manufacturer.

Bioinformatics

Single-pass sequencing of identified cDNA clones of interest was carried out and sequence data analysed using nBLAST and tBLASTx (Altschul et al. 1997, http://www.ncbi.nlm.nih.gov/). Where the deduced EST sequence identities were similar, multiple alignments were carried out using CLUSTALW (Chenna et al. 2003, http://www.ebi.ac.uk/clustalw) to identify and exclude duplicate sequences. Prediction of theoretical pI and other sequence analyses were carried out using http://www.expasy.org. Heat maps for visualisation of expression data were generated using Cluster and Treeview (http://rana.lbl.gov/EisenSoftware.htm) programs (Eisen et al. 1998). EST and cDNA sequences were submitted to the NCBI database (http://www.ncbi.nlm.nih.gov) using dbEST batch submission or Bankit submission protocols respectively, and the individual accession data are given with the corresponding tables.

Results

With the aim of generating a more complete understanding of the processes that occur in the cassava storage root during PPD we have used two complementary microarray hybridisation strategies. Firstly, transcript expression levels were examined over a PPD time course. For each of the five hybridisation experiments the “control” Cy3-labelled target was prepared from root mRNA immediately after harvest and the “experimental” Cy5-labelled target from root mRNA at 12, 24, 48, 72 and 96 h after harvest, respectively. Analysis of these data would be expected to identify genes showing significant up- or down-regulation during PPD. Secondly, we used three microarray hybridisation experiments to compare expression levels between cassava roots and leaves under different conditions. For these experiments mRNA was purified from both roots and leaves at 0 and 24 h after harvest or wounding. A comparative analysis of the hybridisation of the root at 0 h after harvest vs. leaves at 0 h against root at 0 h after harvest vs. leaves at 24 h, together with roots 24 h after harvest vs. leaves 24 h after wounding, would be expected to identify potential root-specific genes, including a subset of PPD-related genes that are expressed in the root prior to harvest but which may subsequently also show differential expression. On the other hand, the comparison of root 24 h after harvest vs. leaves 24 h after wounding alone should allow identification of root injury-specific and thus potentially PPD-specific genes which show de novo induction in the root after injury.

The control genes included in the microarrays served two purposes: (a) as “landing lights” to aid correct grid alignment for the spot detection and measurement steps; and (b) to ensure correct channel allocation. For each output file we confirmed that the control genes showed expected expression and that replicates were reasonably consistent; the controls were then excluded from further analysis (data not shown).

PPD time course microarray analysis

A scatter plot of signal intensities of all 11,136 cassava root cDNA clones for each of the microarray time course experiments is shown in Fig. 1. Each spot represents the mean of 12 normalised signal intensity values (four technical replicates for each of three biological replicates). The majority of transcripts remained unchanged during the post-harvest period, with a peak time-point of differential gene expression observed at 24 h after harvest.

Fig. 1
figure 1

Scatter plot graphs of normalised channel intensities of all ESTs on the PPD microarray for the time-point hybridisation experiments. Signal intensities for each clone are plotted with control (0 h) and experimental (12, 24, 48, 72 or 96 h after harvest) on the X and Y axes, respectively. The diagonal lines indicate 1.5-fold induction/repression ratio cut-offs

Full size image

Based on the selection criteria described in Sect. ”Materials and methods”, transcripts showing differential expression at the 95% confidence level were identified. For the up-regulated genes, only those clones showing a consistent expression profile across the three biological replicate hybridisations were selected for sequencing. In addition, as PPD is an active process it was expected that up-regulated genes would be of greater interest with respect to PPD than down-regulated genes. Thus, for the PPD down-regulated clones, stricter selection criteria were applied; only those showing both a consistent expression profile across the three biological replicate hybridisations and with greater than twofold down-regulation were selected for sequencing.

While no specific effort was made to measure redundancy in the cDNA microarray by sequencing random clones, an impression of this can be gained by comparing the number of clones initially identified in this experiment with that of the non-redundant ESTs, discussed below. The mean representation in the array of each EST was 2.76 with a standard deviation of 3.87. This high variability was due largely to four clones that were represented ten or greater times in the initial selection. It is important to take into consideration that the clones that were selected in this paper were done so using criteria that included significant changes in gene expression and as a result their representation in the microarray may not be wholly comparable to those other clones that do not fulfil these selective criteria.

Based on BLAST homology searches and a review of the literature, putative identities and possible functions were assigned and are shown in Table 1. Discounting duplicated sequences we were able to identify 63 non-redundant ESTs showing up-regulation during the PPD time course and 9 that showed down-regulation. A heat map of expression profiles for these ESTs is shown in Fig. 2. This shows nine ESTs encoding xyloglucan endotransglycolase, ribosomal protein L5, 1-aminocyclopropane-1-carboxylase (ACC) oxidase, pentatricopeptide protein, peroxidase, adenosine triphosphate/adenosine diphosphate (ATP/ADP) translocase, cytochrome b 5, heat shock protein and an unknown protein, showed strong up-regulation within 12 h after harvest. A larger group showed strong up-regulation within 24 h after harvest. Both groups may be considered as early PPD-responsive genes. ESTs corresponding to a catalase MecCAT1 (AF170272; Reilly et al. 2001), a hydroxyproline-rich glycoprotein MecHRGP1 (AF239615; Han et al. 2001), and two Cytochromes P450 sequences CYP79D1 and CYP79D2 (AAF27289 and AAF27290; Andersen et al. 2000), previously shown to be expressed in the storage root after harvest were amongst those identified as showing up-regulation in the microarray.

Table 1 Post-harvest physiological deterioration (PPD)-responsive and putative PPD-specific ESTs
Full size table
Fig. 2
figure 2

Heat map showing the expression profile of 72 non-redundant expressed sequence tags showing differential expression over a PPD time course. Red indicates up-regulation, green indicates down-regulation

Full size image

Reactive oxygen species turnover

It is increasingly apparent that ROS and the enzymes responsible for their interconversion are involved in a range of plant developmental and stress responses including defence, pathogenesis, PCD and senescence. A large minority of the up-regulated genes identified in Table 1 (18%) is involved in the turnover of ROS, and may function directly to detoxify harmful ROS or indirectly to utilise ROS in cellular functions. The catalases (EC 1.11.1.6), ascorbate peroxidases (EC 1.11.1.11) and secretory peroxidases (EC1.11.1.7) catalyse the reduction of H2O2 either directly, using ascorbate or a range of organic or inorganic substrates as an electron donor. MecCAT1 (DT883577, AF170272) was previously isolated in this laboratory and showed up-regulation in the cassava storage root during PPD and during the hypersensitive response (HR) of cassava leaves to the incompatible pathovar Pseudomonas syringae (Kemp et al. 2005; Reilly et al. 2001, 2003). EST DN740367 encodes a novel cassava root secretory peroxidase with similarity to cationic plant peroxidases and a predicted pI of 8.53. Since secretory peroxidases play a role in the vascular streaking reactions of PPD this clone was fully sequenced and submitted to the NCBI database (MecPX3, AY973612). Our data also indicate up-regulation of both a thioredoxin peroxidase (DT883578) and a thioredoxin-like protein (DT883579) during PPD. The thioredoxin peroxidases or peroxiredoxins (EC 1.11.1.15) catalyse the reduction of either H2O2 or various alkyl hydroperoxides to water and the corresponding alcohol in the presence of thioredoxin proteins. EST DT883580 encodes a glutathione-S-transferase (GST); the GSTs catalyse the conjugation of glutathione to a range of electrophilic, hydrophobic and cytotoxic substrates, thereby reducing their toxicity. During oxidative stress they detoxify metabolites resulting from oxidative damage such as lipid peroxidation and oxidative DNA degradation products (Rentel and Knight 2004). ESTs DT883581 and DT883582 encode metallothionein and quinine oxidoreductase sequences, respectively. The metallothioneins are small Cys-rich proteins involved in ROS scavenging and metal homeostasis (Wong et al. 2004). The quinone oxidoreductases serve to break down quinones and semiquinone intermediates resulting from the reduction of quinones. Semiquinones in particular can donate electrons to oxygen, causing the production of superoxide anions (Matvienko et al. 2001). ESTs DN740363 and DT883583 show similarity to plant auxin-induced aldo/keto reductase (AKR) and to early light inducible protein (ELIP), respectively. Both AKR and ELIP play a role in the detoxification of ROS produced during plant stress (Zhang et al. 2005).

Signal transduction and perception

At least three clones identified in this analysis could potentially function in signal transduction or perception events during PPD. ACC oxidase (EC 1.14.17.4) catalyses the last step of ethylene biosynthesis. The plant hormone ethylene is produced in the cassava root within hours after injury (Hirose 1986) and an ACC oxidase sequence (EST DN740375, full length cDNA, MecACCO2, AY973618) with peak expression in the 0 h vs. 12 h microarray hybridisation was identified here. Previous work had identified two other ACC oxidases in cassava, MecACCO1 (EF035079) and MegACCO3 (EF055260), and it is interesting that neither of these members of the gene family were expressed significantly during PPD so as to be identified by the methods used here. ACC oxidase up-regulation in response to wounding has been reported in several plant systems as has differential regulation of ACC oxidase family members (Coupe et al. 2003). EST DN740357 encodes a novel cassava phospholipase. In other plants phospholipases are activated in response to cellular and environmental cues and play a role in signal transduction during stress responses via the formation of lipid-derived messengers, as well as in membrane remodelling and degradation (Wang et al. 2002). EST DN740369 encodes an immunophilin (FK binding-type peptide isomerase). The function of plant FK-binding proteins (FKBPs) is not yet clear; however, they are proposed to be involved in trafficking of signal proteins in plants. Several plant FKBPs are induced by heat and other stress (Krishna and Kanelakis 2003).

Programmed cell death

Previous studies both in this laboratory and elsewhere (Huang et al. 2001) demonstrated up-regulation of genes potentially involved in PCD during PPD. Two novel genes which could potentially function in PCD responses—a cysteine protease (EST DN740364) and a class IV chitinase (EST DN740377) were identified as up-regulated genes in this study. The cysteine proteases are proteolytic enzymes involved in the hydrolysis of proteins and there is increasing evidence that specific proteases may also act as mediators of signal transduction and/or effectors of PCD. They have been implicated in senescence and in response to stresses, such as drought, cold, wounding, ethylene treatment and glucose starvation, which can also induce senescence. PCD may be modulated via the opposing activities of proteases and their cognate inhibitors (Coupe et al. 2003; Alesandrini et al. 2003). The chitinases (EC 3.2.1.14) catalyse hydrolysis of chitin, a major component of fungal cell walls as well as insect exoskeleton. Plant chitinases were originally classified as pathogenesis-related proteins. More recently a role during PCD has been proposed for class IV chitinases in B. napus and Arabidopsis. (Kasprzewska 2003; Lima et al. 2002).

Stress response

Expressed sequence tags DT883584 and DT883585 encode a dehydrin and an Hsp70 sequence, respectively. The dehydrins are classified as group II late embryogenesis abundant (LEA) proteins. They are hydrophilic, thermostable and contain a significant proportion of charged amino acids, implying that they may function as structure stabilisers with detergent- and chaperone-like properties. They are expressed during periods of water stress, or in response to other environmental stresses where osmotic stress is a component of the stress mechanism (Heyen et al. 2002). Plant Hsp70 genes are induced by a variety of stress conditions including heat shock, wounding, water deficit, ABA and cold, and could play a general role in stress adaptation (Li et al. 1994). The water loss during PPD probably imposes water-deficit and osmotic stresses on the root, which would account for the expression of these two genes.

Ion, water or metabolite transport

The ESTs DN740371 and DN740353 encode plasma-membrane intrinsic proteins (PIP), PIP1 and PIP2 type aquaporins, respectively. The aquaporins are membrane channel proteins that regulate water permeability by increasing water permeation across biological membranes. They are members of the major intrinsic protein family. Since stress may alter water regulation, aquaporins have been proposed to be involved in plant stress responses (Baiges et al. 2002). EST DN740350 encodes a gamma adaptin. The adaptins are subunits of adaptor protein complexes, which are involved in intracellular vesicle transport (Lam et al. 2002). EST DN740360 encodes a proton translocating inorganic pyrophosphatase (H+-PPase) (also called V type PPase), a proton pump found in most land plants and in some algae, protozoa, bacteria and archaea. It is primarily localised to the vacuolar membrane of plant cells and hydrolyses pyrophosphate, a by-product of metabolic processes such as protein, starch and cellulose synthesis, coupled to active proton transport across the vacuolar membrane and leading to acidification of the vacuole. Changes in H+-PPase activity occur in response to stress including nutrient stress, anoxia and chilling (Etienne et al. 2002). EST DT883571 encodes an ATP/ADP translocase, an anti-porter of the inner plastid and mitochondrial membrane. ATP is usually exported to the cytoplasm and ADP imported for further ATP synthesis.

Cell wall remodelling and repair

Current models of cell wall structure postulate a pectin matrix, in which is embedded a network of cellulose microfibrills crosslinked to xyloglucans, and a network of structural proteins crosslinked together and to other wall components. Important protein families involved in cell wall metabolism include enzymes for monosaccharide activation and interconversion, polysaccharide synthesis, reassembly and degradation enzymes, structural proteins and glycoprotein glycosyltransferases (Girke et al. 2004). EST DT883569 identified in Table 1 encodes a xyloglucan endotransglycosylase (XET) that shows rapid up-regulation after harvest. The XETs (EC 2.4.1.207) are reassembly and degradation enzymes that function to break and join xyloglucan polymers, the major hemicellulose of the primary cell wall, thereby reinforcing the wall during growth and in response to mechanical strain (Antosiewicz et al. 1997). EST DT883565 encodes a uridine diphosphate (UDP)-glucose dehydrogenase (EC1.1.1.22), an enzyme controlling the biosynthesis of sugar nucleotides required for biosynthesis of hemicellulose components (Tenhaken and Thulke 1996). HRGP/extensin transcripts commonly show up-regulation in response to wounding, pathogen infection, ethylene treatment, and during development. Once synthesised, extensin monomers rapidly become insolubilised in the cell wall in a reaction involving H2O2 and mediated by a cell wall peroxidase (Showalter 1993), thereby sealing and strengthening the cell wall. EST DT883566 described here was identical to a previously described cassava HRGP, cMeHRGP1 (AF239615), which showed up-regulation in cassava cell-suspension cultures treated with glucan cell wall elicitor, and in injured cassava storage roots (Gómez-Vásquez 2000; Han et al. 2001), while EST DT883567 encodes a novel cassava HRGP. Transcript DN740379 encodes a germin-like protein (GLP). The germins and GLPs are a large and diverse class of plant cell wall proteins belonging to the cupin superfamily. They are implicated in cell wall remodelling during both development and stress responses, although the exact nature of their role in these processes is unclear (Mathieu et al. 2003).

Biosynthesis and metabolism

Previous data had indicated that PPD is an active process involving the synthesis of novel compounds as well as active protein synthesis. As might be expected, a number of up-regulated genes involved in biosynthesis and metabolism, and in transcription or translation were identified in this analysis. Five discrete cytochrome sequences showed up-regulation during PPD, a cytochrome b 5 gene and four cytochrome P450 encoding genes. The cytochrome P450 proteins are one of the largest families of proteins in plants with 272 (including 26 pseudogenes) in Arabidopsis. They catalyse a diverse range of reactions, generally based on activation of molecular oxygen with insertion of one its atoms into the substrate and reduction of the other to produce H2O and are classified as mono-oxygenases. The substrates and roles of plant cytochrome P450 s are diverse and include synthesis of structural molecules such as lignin and suberin, synthesis of hormones and signalling molecules, pigments, antioxidants and defence molecules, catabolism of signalling molecules such as abscisic acid (ABA), and detoxification of exogenous molecules such as pesticides and pollutants (Werck-Reichhart et al. 2002). Three of the sequences described here correspond to previously described cassava cytochrome P450 genes. ESTs DN740355 and DN740359 encode cassava CYP79D1 and CYP79D2 cytochrome P450s, respectively, and are involved in the synthesis of the cyanogenic glycosides linamarin and lotaustralin in the root (Andersen et al. 2000). EST DN740365 was identical to cassava cytochrome CYP71E (c15, AY217351). The transcript was previously identified by differential screening to identify cassava storage-root-specific promoters, and was expressed in storage roots but not leaves. Sequencing and analysis of the cognate promoter identified a number of cis-acting regulatory motifs including a wound-responsive motif, an ethylene response element, an ABA-responsive element, and two methyl jasmonate response elements (Zhang et al. 2003). EST DN740376 encodes a novel cassava cytochrome P450, whilst EST DT883570 encodes a novel cassava cytochrome b 5. In yeast and mammalian cells cytochrome b 5 acts as a redox intermediate in biosynthetic reactions. In plants it is proposed to act as a direct or indirect electron donor in the biosynthesis of products such as sterols, terpenoids and gibberellic acids (Martsinkovskaya et al. 1999). ESTs DN740366 and DT883564 encode two discrete UDP glycosyltransferases. The glycosyltransferases (EC 2.4.x.y) are responsible for the biosynthesis of disaccharides, oligosaccharides and polysaccharides and catalyse the formation of glycosidic bonds. Such glycosylation reactions often represent the end point of higher plant secondary metabolite synthesis and, in addition, many classes of plant hormone exist in a glycosylated form including auxins, gibberellins, cytokinins and ABA (Lim et al. 2003). The sequences presented here showed no significant sequence similarity to previously isolated cassava glycosyltransferases available on the NCBI database (Hughes and Hughes 1994). EST DT883572 encodes the enzyme l-asparaginase (EC 3.5.1.1), a key enzyme involved in Asn hydrolysis in plants. Ammonia is released in the reaction and is utilised for the synthesis of nitrogen containing compounds, including amino acids required for protein synthesis (Casado et al. 1995). EST DT883574 encodes a ketol-acid reductoisomerase (EC 1.1.1.86), an enzyme involved in synthesis of branched chain amino acids including isoleucine and valine (Dumas et al. 1993). EST DT883573 encodes transaldolase (EC 2.2.1.2), an enzyme of the non-oxidative branch of the plant pentose phosphate pathway that is involved in transfer of carbohydrates from primary to secondary metabolism allowing synthesis of aromatic amino acids, flavonoids and lignin (Caillau and Quick 2005). EST DT883575 encodes an arginine decarboxylase (ADC, EC 4.1.1.19), involved in the synthesis of putrescine. Transient increases in ADC activity and putrescine levels have been described in a number of plant stress and wounding responses (Perez-Amador et al. 2002). EST 883576 encodes sulphite reductase (EC 1.8.7.1) a key enzyme of sulphate assimilation in plants (Bork et al. 1998).

Transcription and translation

Eight of the up-regulated clones described in Table 1 code for elements of the protein translation machinery: ribosomal proteins L5, S10, S16 and S29; eukaryotic initiation factor eIF4 involved in mRNA recruitment; an ATP-dependent RNA-helicase involved in unwinding of mRNA secondary structure; a seryl tRNA synthase involved in definition of codon identity; and an elongation factor eF1α involved in binding of aminoacyl tRNA to ribosomes during the elongation step of translation. EST DN740356 encodes an oligouridylate binding protein (UBP) with highest similarity to Nicotiana plumbaginifolia UBP1. UBP1 enhances accumulation of certain classes of transcript via enhanced mRNA stability (Lambermon et al. 2000). EST DT883591 encodes an alternative splicing factor SF2a. Alternative splicing profiles are affected by environmental stress conditions, and are thought to be mediated both by transcriptional regulation of splicing factors and alternative splicing of the splicing factors themselves (Iida et al. 2004). EST DT883590 encodes a pentatricopeptide repeat (PPR) protein. PPR proteins contain a characteristic 35 amino acid motif repeated in a tandem array. Most plant PPR proteins are targeted to mitochondria or chloroplasts and are thought to be involved in RNA processing or translation (Kazama and Toriyama 2003).

Confirmation of expression profiles of selected up-regulated clones

To validate and extend the microarray data a number of cDNA clones corresponding to up-regulated ESTs from the microarray analysis were fully sequenced and used as probes for Northern blotting experiments (Fig. 3). Fully sequenced clones include the cognate cDNAs to ESTs for ascorbate peroxidase MecAPX2 (AY973722), a putative auxin-induced protein aldo/keto reductase MecAKR (AY973615), a catalase MecCAT2 (AY973614), a cysteine protease MecCP1 (AY973616), a germin-like protein MecGLP (AY973617) and a secretory peroxidase MecPX3 (AY973612). Northern blot data confirmed that the expression patterns of the corresponding genes were up-regulated during a PPD time course. The precise timing of peak expression observed was not identical to the microarray data (Table 1; Fig. 2). However, it had previously been observed that the symptoms of PPD initiate and progress more rapidly under field conditions in Colombia than under laboratory conditions at Bath, and this may account for earlier observed peaks of expression in the microarray data, which used targets derived from mRNA isolated from field-grown roots. The observed expression profiles in different cassava tissues indicates that at least some of the genes are potentially PPD-specific, i.e. expressed only during root deterioration and not in un-wounded root; or wounded or un-wounded leaves. Of particular interest were the germin-like protein (MecGLP) and the secretory peroxidase (MecPX3), which showed putative PPD-specificity. The latter was induced to high levels within the 24 first hours of harvest according to both microarray and Northern blotting data. The ascorbate peroxidase (MecAPX2) showed contrasting expression patterns in wounded leaves, being expressed when the wounded leaf remained attached to the plant but not when it was detached. The cysteine protease (MecCP1) showed the opposite expression pattern in leaves, with up-regulation in wounded detached leaves, but background expression in wounded attached leaves. The nucleotide sequence of MecCPI had 80% identity with BOCP3 – a senescence associated cysteine proteases in broccoli florets (Coupe et al. 2003). Injured cassava roots which remain attached to the plant are capable of normal wound healing (Mwenje et al. 1998), thus it is tempting to speculate that severance of the root or leaf from the plant triggers components of a senescence or PCD pathway.

Fig. 3
figure 3

Expression profiles of selected genes. Northern blots of cassava mRNA to ascertain the expression profiles of the genes corresponding to selected cDNA probes over a PPD time course (time in hours) and in leaves (A wounded but attached leaves, D wounded and detached leaves, CK control, un-wounded attached leaves)

Full size image

PPD down-regulated genes

Sequences identified as showing down-regulation in the storage root during PPD include a cystatin like protein (EST DT883603), a LEA protein (EST DT883604), a translationally controlled tumour protein (TCTP) (DT883604), two clones with homology to auxin repressed proteins (ARP) designated ARP1 and ARP2 (DN740380 and DN740381), and a PWWP domain protein (DN740380).

Plant cysteine protease inhibitors or cystatins regulate endogenous cysteine protease activity during various biological processes including defence, response to biotic and abiotic stress, embryogenesis and seed germination, and in PCD (Shyu et al. 2004). Generation of ROS plays a key role in PCD in both plant and animal cells; at low levels antioxidant enzymes are induced, though when ROS levels reach a critical threshold a PCD pathway is triggered. For example, in cultured soybean cells PCD-inducing oxidative stress triggered expression of a set of cysteine proteases, while ectopic expression of an endogenous cystatin inhibited cysteine protease activity and blocked PCD (Solomon et al. 1999).

The LEA proteins are a group of loosely related proteins that are involved in desiccation tolerance in plants. Although their precise mode of action remains unclear proposed mechanisms include water retention, ion sequestration, direct protection of proteins or membranes and renaturation of unfolded proteins. The down-regulated EST DT883604 described here showed homology to LEA5/D73 type LEAs. These have recently been reclassified as class V LEAs (Wise 2003).

Translationally controlled tumour proteins are tubulin-binding proteins that associate with microtubules in a cell-cycle-dependant manner and play a role in cell growth-related activities. They have been primarily studied in mammalian systems and their precise role in plants remains unclear. In pea a TCTP (PsCaP23) and a 60S ribosomal protein L41 were up-regulated within 15 min of root cap induction, where their expression was localised to dividing cells within the root cap, suggesting that plant TCTPs, like their mammalian counterparts, are involved in cell proliferation (Woo and Hawes 1997). In humans a TCTP is involved in the modulation of translation elongation, via interaction with the elongation factors eEF1α and eEF1β. Decreased TCTP expression levels promote apoptosis, and strong down-regulation has been observed during tumour reversion/suppression (Cans et al. 2003).

The ARP gene family is specific to higher plants, and members of the family contain highly conserved N-terminal and C-terminal domains. The plant hormone auxin is involved in the regulation of a range of growth and developmental processes; however, to date little is known about the precise role of ARP genes. An ARP homologue from Capsicum annuum was induced in X. campestris and P. capsici infected tissues but was not induced in response to wounding, salicylic acid, jasmonic acid or ethephon treatment (Jung and Hwang 2000). The Robinia pseudoacacia RpARP gene was down-regulated by auxin, and such down-regulation was dependant on the presence of metabolisable sugar and on protein synthesis. Deletion constructs of RpARP with or without untranslated regions (UTRs) indicated that the protein was additionally regulated post-transcriptionally by auxin and this is mediated via the UTRs. Although a number of mRNA instability motifs have been identified in the UTRs of other genes, no known motif was found in RpARP (Park and Han 2003). Intriguingly, although the available coding regions of cassava ARP1 and ARP2 are identical, the full sequences of the cognate cDNAs reveal a difference in their 3′ UTRs with a ∼40bp insertion in ARP1 (AY973620) relative to ARP2 (AY973621).

EST DN740380 showed highest similarity to an Arabidopsis PWWP domain protein. The PWWP domain is a conserved protein domain based on a central core “proline–tryptophan–tryptophan–proline”, hence its designation (pfam PF00855). The domain is located at either the N- or C-terminus of the protein and is found in proteins involved in cell division, growth and differentiation. Where data are available, PWWP proteins appear to be nuclear proteins and are often DNA binding proteins acting as transcription factors. It is proposed that the domain acts as a site for protein–protein interaction and may modulate transcription via chromatin remodelling (Stec et al. 2000).

Microarray analysis to identify PPD-specific and storage-root-specific genes

In addition to genes that were up- or down-regulated during PPD, the microarray hybridisations and analyses were capable of identifying genes whose expression change significantly (up or down) during PPD (storage root wounding) but not during leaf wounding, and genes expressed in storage roots but not in leaves (un-wounded or wounded). While not categorical definitions, the former can be considered as PPD-specific, or perhaps more accurately storage-root-wounding-specific, and the latter root-specific. Twenty four storage-root-wounding-specific genes were identified and are indicated in Table 1. They were distributed between the all the several functional classes of up- and down-regulated PPD genes discussed above (Table 1). Surprisingly, while this storage-root-wounding-specific group included the secretory peroxidase (MecPX3), also identified by Northern hybridisation (Fig. 3) and discussed above, the germin-like protein (MecGLP) did not fulfil the microarray criteria.

Of the group of 21 storage root-specific genes, none were common to the up-regulated PPD genes, though four were identical to the smaller group of down-regulated genes: cystatin-like protein (DT883603), TCTP (DT883602), ARP1 (DN740380) and an unidentified expressed protein (DN740383) (Table 2; Fig. 4). Interestingly, the first two were also common to the PPD-specific group. The remaining 17 genes in this root-specific group were novel and fell into several functional groups (Table 2). Of these, EC591287 was very highly similar to an allergenic-related protein from cassava (AY101376, de Souza et al. 2004) and also bears high similarity to another cassava gene for a glutamic acid-rich protein (AY217354) whose promoter shows strong expression in vascular tissues and during the secondary growth of cassava storage roots (Zhang et al. 2003).

Table 2 Root-specific ESTs
Full size table
Fig. 4
figure 4

Heat map showing the expression profiles of 21 non-redundant expressed sequence tags showing differential expression in deteriorating cassava storage roots and leaves. Red indicates up-regulation, green indicates down-regulation

Full size image

Discussion

The approach employed here has led to the identification of 73 genes whose expression changed significantly during PPD. In addition, a comparison between harvested storage roots and un-wounded and wounded leaves led to the identification of a 24 member subset of these genes as being storage-root-wounding-specific and another 21 genes, 4 of which were included in the PPD down-regulated group, as storage-root-specific. It is reassuring that many of the genes studied previously as “likely candidates” involved in PPD have also emerged in this more rigorous and “operator neutral” microarray approach (Han et al. 2001; Reilly et al. 2001, 2003).

A recent examination of the cassava fibrous and storage root proteome led to the identification of 237 protein spots of which 104 were listed (Sheffield et al. 2006). The classification of these proteins into functional groups placed them in many of the classes identified in this study. However, only six proteins, ascorbate peroxidase, catalase CAT1, GST, EF1α, transaldolase and TCTP, were common to both. The EF1α was identified here as up-regulated during PPD and storage-root-wounding-specific, while the TCTP was down-regulated, storage-root-wounding-specific and root-specific.

Huang et al. (2001) used a cDNA-ALFP fingerprinting approach to identify transcript-derived fragments (TDFs) of genes whose expression altered during cassava PPD; these included TDFs for genes involved in oxygen scavenging and stress responses. To our knowledge their TDF sequences are not available on public databases, thereby preventing direct sequence comparisons. However, a peroxidase, glycosyltransferase, elongation factor EF1α, heatshock protein 70, three ribosomal proteins, and two cytochrome P450 sequences were shown to be up-regulated in both Huang et al. (2001) and this study. While no comparable microarray study of PPD exists, that of Lopez et al. (2005), using a microarray of cassava ESTs, enriched for starch biosynthetic and bacterial blight-responsive sequences, to dissect cassava’s response to the pathogen X. axonopodis, merits comparative examination. Both studies identified some common functional groups of genes whose expression was up-regulated: ROS modulation, cell wall metabolism, and transcription/translation. For example, both identify several ESTs for ROS-modulating enzymes that fall into essentially the same classes or gene families; however, a close sequence comparison shows that only one pair of ESTs for ascorbate peroxidase (DN740373 and CK649432), were essentially identical with respect to the regions sequenced (data not shown). Other common genes include cytochrome P450s, of which DN740367 and CK63363 are identical (data not shown), and aquaporins. These common genes confirm that, while the biotic and abiotic stress responses analysed in these two studies are strikingly different, there are major areas of overlap between plant stress responses in general.

The characterisation of genes expressed during PPD and isolated through random and directed cloning approaches had led to the generation of a model that placed ROS and their modulation centre-stage in the PPD response, supported by enzymes involved in PCD and more general defence responses (Reilly et al. 2003). The data presented in this paper, which have been collected by a more thorough and statistically robust approach, provide further support for that model. The ROS produced and triggered by the wounding associated with cassava root harvesting probably fulfil multiple roles: as initial signals that trigger signal transduction pathways that lead to changes in gene expression; as components in cellular development or repair, such as cell wall formation; as oxidants of cellular components and molecules, such as of scopoletin that leads to the vascular streaking of PPD; and cellular damage or PCD. All these ROS-related events were observed during PPD and increased activity of genes involved in several of these processes was detected during this study, confirming an important role for ROS in the development of PPD.

The expression of some genes associated with PCD or apoptosis were significantly altered: cysteine protease (MecCP1, AY973616) and class IV chitinase (DN740377) were up-regulated, whilst the cystatin-like protein (DT883603) and TCTP (DT883602), were down-regulated. MecCP1 was up-regulated in wounded detached leaves but not in wounded attached leaves, suggesting that the former had embarked on a PCD pathway while the latter had not. Cystatins regulate cysteine protease inhibitors, so the up-regulation of the former coupled with the down-regulation of the latter would be expected to enhance the hydrolysis of proteins and thereby the PCD response of the root. TCTPs are highly conserved proteins that appear to be common to all eukaryotes; they are important in cell cycle progression and have anti-apoptotic activity. While TCTP is not tumour-specific, its expression increases during mammalian tumour formation and its down-regulation is associated with reversion of cells to a normal phenotype (Bommer and Thiele 2004). The function of TCTP has not been studied in plants; however, its down-expression observed during PPD is consistent with the anti-apoptotic role observed in animal systems. Certainly, the down-regulation of two anti-apoptotic genes and the parallel up-regulation of two PCD genes is consistent with a role for PCD responses during PPD. Additional circumstantial data supporting this view comes from the observation that the coumarin, scopoletin, that accumulates during PPD and is responsible for much of the fluorescence under UV light (Buschmann et al. 2000b), accelerates the PPD response when applied exogenously to the cassava root (Wheatley and Schwabe 1985). Recently it has been reported that scopoletin has an inhibitory effect and induces apoptosis in human cancer cell lines (Kim et al. 2005; Adams et al. 2006). These reports raise the intriguing question as to whether scopoletin may also have a pro-apoptotic role in cassava PPD.

These data are not sufficient to determine whether in the harvested cassava storage root the accumulation of high levels of ROS is a symptom or component of PCD, or vice versa. It is possible that the rapid oxidative burst detected within 15 min of harvest (Reilly et al. 2003) generates the initial signals that trigger PCD of which high levels of damaging ROS are a major component. Nonetheless, the results presented here provide knowledge and tools with which to approach answering these questions through modulating the root’s anti-oxidant or anti-PCD status via transgenes driven by specific promoters.