Comparative transcriptome analysis coupled to X-ray CT reveals sucrose supply and growth velocity as major determinants of potato tuber starch biosynthesis

Starch biosynthesis in potato leaves follow carbohydrate accumulation and show similarities to tuber starch biosynthesis

Leaf samples for starch and RNA extractions were taken from potato plants grown under a 14 hour light and 10 hour dark cycle. The light conditions were chosen to ensure diurnal turnover of transitory starch. As was expected, starch levels were the lowest at the start of the light period and accumulated during the day to a high at the end of the light period (Figure 2A). The same pattern was observed for sucrose although the accumulation and decline was more rapid (Figure 2B). The levels of starch and sucrose compared well to what was previously measured in potato leaves [4].

After confirming that there was a turnover of transient starch under the specific light conditions used, gene expression at different times of the day was analysed by microarray. To investigate how much similarity there is between leaf and tuber starch biosynthesis, the expression of genes known to be important for tuber starch biosynthesis was investigated in leaves. Starch biosynthetic genes had a strong diurnal rhythm of expression. Interestingly, genes involved in the import of glucose 6-phosphate and ATP into the plastid are co-regulated with ADP-glucose pyrophosphorylase, suggesting that both genes are also involved in transitory starch biosynthesis in leaves (Figure 3A). To rule out that the observed expression of GPT and NTT in leaf extracts was due to contaminating epidermis rather than mesophyll cells, cell-specific RNA analysis was performed. To this end epidermis stripes were harvested and expression of GPT and NTT were probed by quantitative real-time PCR (qRT-PCR). As shown in figure 4, GPT and NTT expression in epidermis cells was about ten fold lower compared to whole leaf extracts. This supports the hypothesis that GPT and NTT are indeed required for photosynthetic starch biosynthesis in leaves. However, other cell types such as companion and bundle sheath cells could still contribute to the observed expression pattern. It can also not be excluded that the transporters are exporting substrates from the plastid.

Most starch synthases and branching enzymes had a similar pattern to the above mentioned genes in figure 3A with the exception of GBSS which was highest expressed two hours into the light period (Figure 3B). Smith et al. [28] argue and provide evidence that since the enzyme is present within the granule, the protein is degraded together with starch at night and must very quickly be re-synthesised in the morning. Two genes involved in the light reaction of photosynthesis, plastocyanin and chlorophyll a/b binding protein, were already up-regulated at the first time point which was taken moments after the lights came on and were highest expressed two hours into the light. Ribulose 1,5 bisphosphate carboxylase oxygenase (Rubisco) had an expression pattern similar to that of starch biosynthetic genes (Figure 3C). The sucrose cleavage enzymes cell wall-bound invertase and sucrose synthase had very different expression patterns with Susy 4 being much stronger regulated. Susy 4 increased during the light and went down in the dark. Hexokinase 1 was not diurnally regulated, whilst hexokinase 2 and fructokinase had a similar pattern of increasing early in the morning and declining at the end of the light period (Figure 3D).

For analysis of gene expression during tuber induction, tuber developmental stages were designated according to Kloosterman et al. [26] and the expression of genes were studied by microarray analysis. Stage 1, representing an unswollen stolon, showed very low expression for all starch biosynthetic genes except starch synthase IV which was highest expressed at stage 1, fitting with its proposed role in starch granule initiation [22, 23]. From stages 3-5 there was an increase in the expression of all starch biosynthetic genes with the exception of starch synthase IV (Figure 5A and 5B). As evidence that stage 1 was before the onset of tuberisation, cell wall-bound invertase expression was very high and sucrose synthase 4, a tuber expressed isoform, low. From stages 3-5 the expression cell wall-bound invertase went down and that of Susy 4 increased (Figure 5C). The expression patterns compared well to earlier experiments [29]

qRT-PCR is a well accepted method for verifying microarray data and was used to validate the expression patterns of a selected number of genes in independent samples. The expression patterns and levels of Susy 4, GPT and GBSS compared well to the microarray data in both leaves and tubers and confirmed the accuracy of the microarray results (Figure 6).

In contrast to starch biosynthesis, the pathway of starch degradation in potato tubers is not well understood. Currently it is believed that the major pathway of starch degradation in Arabidopsis and other organisms proceeds via ί-amylase [32]. Starch must first be phosphorylated by glucan-water dikinase (GWD), and to a lesser extent phospho-glucan water dikinase [33, 34]. It is then either directly degraded to maltose by ί-amylase or first to linear glucans by Isoamylase 3. Before complete hydrolysis of phosphorylated glucans can occur, phosphate groups must be removed by specific a phosphatase, known as SEX4 phosphoglucan phosphatase [35]. The linear glucans are then hydrolysed to maltose by ί-amylase. Maltose is subsequently exported to the cytosol via the maltose transporter (MEX1) [36] and it is believed that the disproportioning enzyme 2 (DPE2) plays an important role in its further degradation [37]. Maltose can also be degraded by disproportioning enzyme 1 (DPE1) inside the plastid. For potato evidence has been provided that DPE2 might be present inside the plastid [38], which questions the role of MEX1. If the expression data is compared to that of Arabidopsis, indications are there that a similar pathway for starch degradation probably occurs in potato leaves, except that isoamylase 3 was not differentially expressed. The other genes mentioned were all differentially co-expressed in leaves, and expression patterns followed that of starch accumulation in the leaf (Figure 7). The entire microarray dataset for both the diurnal leaf experiment and the tuber induction experiment have been deposited on ArrayExpress (accession numbers E-MEXP-2481 Ferreira et al. Potato diurnal leaf time-course and E-MEXP-2482 Ferreira et al. Potato tuber induction). POCI sequence and annotation data are available through the POCI online tool http://pgrc.ipk-gatersleben.de/poci.

Diurnal oscillation of GBSS in potato tubers can be linked to differences in sucrose supply

It is known that starch biosynthetic genes are diurnally regulated by several factors, with sucrose and the circadian clock seemingly being the most important. Bläsing et al. [39] showed that between 30-50% of genes in Arabidopsis rosettes show diurnal changes in their transcripts and that this was especially true for genes involved in redox regulation, nutrient acquisition and assimilation, and starch and sucrose metabolism. Comparative analysis of nutrient feeding and diurnal transcription profiles indicate that sugars make a major contribution to diurnal regulation. Furthermore, Osuna et al. [40] analysed gene expression in carbon-deprived Arabidopsis seedlings after the addition of sucrose. Genes involved in central carbon metabolism, and more specifically starch biosynthesis, showed a response to sucrose and this did lead to an increase in starch content. A second major regulator of diurnal gene expression seems to be the circadian clock. Bläsing et al. [39] identified a subset of 373 genes known to be circadian regulated [41]. The gene set was used in a principle component analysis which showed that sucrose and the circadian clock are the predominant factors in regulating diurnal gene expression and that light, nitrogen and water deficiency makes a smaller contribution. Starch degradation related genes also have a strong diurnal rhythm in Arabidopsis which was maintained under continuous light, but not continuous darkness [42].

Since GBSS is one of the strongest diurnally regulated genes involved in starch biosynthesis [28], it was investigated whether this gene is also diurnally regulated in tubers. GBSS expression was measured at different times of the day and under different light regimes. Plants were grown in normal long day conditions and samples were taken from stolon-ends and tubers at different time-points of the diurnal cycle. GBSS expression in tubers oscillated during the day and was highest at the end of the light period and lowest 2 hours into the next light period (Figure 8A). Expression was also significantly lower in tubers from plants kept in twenty four hours of darkness compared to tubers from plants kept in a normal light/dark cycle (Figure 8B).

As mentioned, sucrose and the circadian clock seem to be important regulators of diurnal gene expression and this is especially true for GBSS. Tenioro et al. [43] showed that GBSS is strongly regulated by the circadian clock and that expression is markedly lower in mutants lacking clock genes LHY and CCA-OX respectively. In a detailed study done on GBSS in snapdragon (Antirrhinum majus) it was shown that GBSS is diurnally regulated in leaves even under continuous light and it was concluded that the regulation is due to the circadian clock. This was not the case in snapdragon roots though, where expression was the same in the middle of the day and in the middle of the night [44]. Also in rice leaves GBSS continues its diurnal cycling under continuous light suggesting circadian regulation, but expression can be induced by nitrogen starvation or sucrose feeding and repressed by darkness, indicating the importance of sucrose in its regulation [45]. Moreover, sucrose floating experiments with potato leaves show that GBSS expression can be induced by sucrose [46].

To investigate whether oscillation of GBSS might be due to changes in sucrose supply from the source, sucrose content of stolon-ends attached to a tuber was measured. Phloem sucrose content is very high compared to those tissues surrounding it [47] and phloem signifies a large proportion of the total stolon tissue [48]. This makes it possible to determine sucrose import to the tuber by measuring sucrose content of the stolon-end [3]. Sucrose content differed significantly during the day and was highest at the end of the light and lowest at the end of the dark period. When plants were kept in constant darkness, sucrose content of stolon-ends declined linearly over time indicating that changes in sucrose supply from the leaves contributes to the oscillating expression of GBSS in tubers (Figure 8C).

Although GBSS expression in tubers can be linked to diurnal changes in sucrose supply from the source, caution should be exercised in the interpretation of the significance of this in terms of enzyme activity. As mentioned earlier, the level of GBSS protein does change substantially during the day in Arabidopsis leaves and the reason for this is probably the location of GBSS within the granule [28]. This is also true for algae Chlamydomonas reinhardtii suspension cultures where GBSS expression and enzyme activity correlates with starch levels. The authors state that the correlation is probably due to the fact that the analysis was conducted in suspension cultures, where new cells are produced constantly leading to continuous production of new GBSS protein [49]. This however is not necessarily true for GBSS in other tissues or for other enzymes of starch biosynthesis and several studies have shown that diurnal changes in expression do not lead to changes in protein levels [9, 42]. However, it still remains an interesting finding that GBSS expression in tubers follows a diurnal rhythm which declines when sucrose supply from the leaves are reduced.

Starch biosynthetic gene expression is influenced by tuber growth velocity

As mentioned earlier, tuber initiation and growth rates are not synchronised. Because analysis of tubers growing at different velocities would provide important information on factors determining growth, it was thought necessary to develop a method to determine growth velocity before harvesting the tubers. For this, X-ray CT was used to determine tuber volume over a time course and then calculate the growth velocity. An overview of the method for calculating tuber volume using X-ray CT is given in figure 9. The accuracy of the growth velocity determination depends largely on the accuracy of volume calculation and it was crucial to establish whether tuber volume could be accurately calculated. Figure 10A illustrate that potato tubers were clearly distinguishable from the surrounding soil, and a histogram demonstrated that segmentation of tuber material was possible. The software used for segmentation also allowed for manual correction of possible mistakes made during the automatic segmentation process. Tuber growth velocities were calculated for seventeen tubers. To validate the method, calculated tuber volumes were compared to measured tuber volumes after harvest. X-ray CT calculated volumes showed a correlation coefficient with experimental volume measurements of 0.986 showing that X-ray CT calculated volumes were accurate (Figure 10B).

Six tubers with different growth velocities (Figure 11) were selected for microarray hybridisation and the gene expression patterns were analysed. Interestingly no linear correlation between the expression of starch biosynthetic genes and growth velocity could be observed. The expression of genes involved in starch biosynthesis is shown in figure 12. The expression of these genes, and also other genes involved in starch biosynthesis, did not show large differences in expression levels between tubers that were still growing, albeit at very different velocities. Starch biosynthetic gene expression was however much lower in tubers that have virtually stopped growing and it seems as though the relationship between growth velocity and starch biosynthetic gene expression is qualitative rather than quantitative. This was especially true for Susy, a major determinant of tuber sink strength [11].

These data indicate that tubers that look visually similar have large differences in gene expression depending on their growth stage. Moreover, transcription profiles reveal many genes that are differentially expressed between growing and non-growing tubers which provide important information towards the identification of factors determining tuber growth. The entire microarray dataset for the tuber growth velocity experiment has been deposited on ArrayExpress (accession number E-MEXP-2484 Ferreira et al. tuber growth velocity).

Comparative analysis of transcription profiles reveals genes possibly regulating starch biosynthesis

The similarities observed in terms of starch biosynthetic gene expression between leaves and tubers suggests that the same regulators might be involved in both processes. In an attempt to identify these regulators, comparative analyses of various transcription profiles were conducted. Firstly, features differentially expressed between a stolon/swollen stolon, diurnally in leaves and between growing/non-growing tubers were selected. To further decrease the number of features, only those features differentially expressed under all conditions were selected which reduced the number to 1662 (Figure 13A). K-means clustering analysis was performed on these features and most starch biosynthetic genes were present in three clusters. These clusters also showed the expected expression patterns for starch biosynthetic genes (Figure 13B). A functional assignment on the 913 features present in these three clusters was performed which showed that 37% of the features were involved in metabolism and a further 9% storage protein related (Figure 13C). Transcription factors made up 6% of the features and these were further investigated.

Arabidopsis orthologs for these transcription factors were identified and used to gather more information on their possible function. Six of these transcription factors showed increased expression in Arabidopsis suspension culture fed with sucrose [50] and most of these genes were also down-regulated in leaves when the dark period was extended [51]. Figure 14 contain the expression profiles of these transcription factors including the identifiers of their orthologs in Arabidopsis. All of these transcription factors, with the exception of AT4G00870 and AT4G32730, were also down-regulated in the 35S::amiR-white-1(MIR172a) knockout [52]. The construct targets the GUN4 allele which is important for plastid to nucleus signalling and when this is interrupted it affects plastidic expression of several nuclear encoded genes. Furthermore, the expression of GPT2, plastidial PGM, GBSS and SBEB was also significantly down in these plants. This, along with the fact that these transcription factors were co-expressed in the same clusters as starch biosynthetic genes, indicate that they might be important in regulating starch accumulation.

Plants and growth conditions

Solanum tuberosum (cv Solara) were propagated in tissue culture on MS medium [53] containing 2% sucrose. To obtain tubers, plants were transferred to soil and cultivated until harvest in individual pots in the greenhouse or growth chambers. For transcriptional analysis of potato leaves over a diurnal period, plants were grown in a growth chamber under a 14 hour light and 10 hour dark cycle. Plants for the tuber induction and X-ray CT studies were grown in the greenhouse under normal long day conditions. Plants used for analysing for the diurnal rhythm in tuber gene expression were grown in growth chambers under normal long day conditions.

RNA isolation, cRNA synthesis and Cy3-labeling

Isolation of total RNA was performed as described previously [54].

POCI array and database

The construction of the POCI array and database has been described in detail [29].

Sample preparation and microarray hybridization

For each hybridisation at least two biological replicates were included. For diurnal leaf time-points five different leaves from five plants were sampled for each biological replicate. For tuber induction samples stolon or tuber material from 40 plants were pooled into two groups according to the developmental stages. For the hybridization comparing fast and slow growing tubers, two independent samples were taken from each tuber and treated as biological replicates. RNA purity was measured by the ND-1000 Spectrophotometer (NanoDrop Technologies). To check for RNA degradation two μg of total RNA were separated on 1.5% formaldehyde containing agarose gel. Total RNA was purified using RNeasy Mini Spin Columns (QIAGEN, Valencia, CA). Afterwards RNA quality and quantity was tested using the Agilent 2100 BioAnalyzer (vB.02.03 BSI307) as recommended by manufacturer's protocol (Agilent RNA 6000 Nano Assay Protocol2). Synthesis of cDNA and cRNA was performed as described in the one-color microarray-based gene expression analysis protocol provided by Agilent including the one-color RNA spike-in kit (v5.0.1, 2006; Agilent Technologies, Santa Clara). After fragmentation Cy3-labelled samples were loaded on the array and hybridised over night (17h/65°C). Slides were washed as recommended in the manufacturer's protocol and scanned on the Agilent Microarray Scanner with extended dynamic range (XDR) at high resolution (5 μm). Data sets were extracted by using the feature extraction software (v9.5.3.1/Agilent Technologies) using a standard protocol.

Array data analysis

Data were imported into GeneSpring GX 7.3.1 (Silicon Genetics, Palo Alto, CA, USA) and additionally stored on a local server. A three step normalization was applied: (1) values less than five were set to five, (2) per chip normalization to 50^th percentile and (3) subsequently the signal for each feature was normalised to the median of its value across the entire dataset. For comparative analysis of various transcription profiles, a volcano plot was applied to select for features more than two-fold differentially expressed between two conditions including the Benjamini-Hochberg multiple test correction for the four replicate experiment comparing slow and fast growing tubers. No multiple test correction was employed for the experiments comparing stolon and swollen stolon, and the diurnal leaf experiment, where only two replicates per data-point were used. Features commonly differentially expressed in all experiments were identified using a Venn diagram. K-means clustering was performed using Pearson correlation to split the selected features into five clusters from where features were chosen to conduct a functional assignment.

Starch and sucrose measurements

Sucrose and starch were measured according to a modified method of Müller-Röber et al [17]. To determine sucrose supply rate to a tuber, sucrose content of stolon-ends was determined. A stolon-end is defined as the 15 millimetres of a stolon directly above the connection to the tuber.

qPCR analysis

For all qPCR analysis at least three biological repeats were used unless stated otherwise. Expression levels of genes were determined by real-time quantitative RT-PCR and the corresponding primers for the amplification of targets between 75 and 150 bp were designed using Primer3plus software [55]. Total RNA (five μg) from each of the developmental time points was treated with DNaseI (Fermentas GmbH) before undergoing reverse transcription using oligo d(T) primers and RevertAid™ H minus first strand cDNA synthesis kit (Fermentas GmbH) to generate a first strand cDNA template. Potato ubiquitin primers were used as a control as described previously [26]. One μl of 1:10 diluted cDNA for each time point were amplified with gene-specific primers in three technical replicates on a Mx3000P Q-PCR system (Stratagene) in combination with the Brilliant II SYBR Green Q-PCR Master Mix Kit (Stratagene). The thermal profile was as follows: 1 cycle 10 min at 95°C for DNA polymerase activation followed by 35 cycles of 10 s at 95°C, 15 s 60°C and 20 s 72°C. The primer sequences were as follows: ubi3 (L22576) forward primer, 5'-TTCCGACACCAT CGACAATGT-3'; reverse primer, 5'-CGACCATCCTCAAGCTGCTT-3'. For GBSS the primer sequence was based on POCI feature micro.920.c2. The forward primer was designated GBSS_920.c2 F (5' - CAGACTTGAGGAGCAGAAAGG - 3') and the reverse primer GBSS_920.c2 R (5' - GTGAGCCAAAGGGACATTGA - 3'). For GPT the primer sequence was based on POCI feature micro.1076.c1. The forward primer was designated GPT_1076.c1 F (5' -CCTTGTTTCCTGTTGCTGTG- 3') and the reverse primer GBSS_1076.c1 R (5' -AAAGCAGGCTCTCCACTCTT- 3'). For Susy the primer sequence was based on POCI feature micro.196.c8. The forward primer was designated Susy_196.c8 F (5' -CTGCTGTTTATGGGTTCTGG- 3') and the reverse primer Susy_196.c1 R (5' -GGCACACCTTCATTCACTCA- 3'). For NTT the primer sequence was based on POCI feature micro.1831.c2. The forward primer was designated NTT_1831.c2 F (5' -GAGCAGCAGCCAAGATAACAC- 3') and the reverse primer GBSS_1831.c2 R (5' -GTTCTGCATTGCACCCACA- 3'). Relative gene expression was calculated using the Pfaffl method [56].

Description of X-ray CT

With X-ray CT the 3D volume information of objects can be reconstructed using X-ray projections of the object from different aspects. The geometry used for the investigation was the axial 3D-CT, where a conical X-ray beam projects the object onto a flat 2D image detector. Using axial 3D-CT, projections of the object are taken under different viewing angles, rotating the object perpendicular to the central X-ray beam. The reconstructed volume data set consists of volumetric elements, called voxels, containing grey levels which represent information about the X-ray attenuation characteristics depending on the mass attenuation coefficient and the density distribution of the material [57, 58]. The mass attenuation coefficient itself is dependent on the applied X-ray spectrum and the effective atomic number of the X-rayed material. The calculation of tuber volumes of potted potato plants embedded in soil requires the segmentation of tubers from other materials in the X-ray CT volume data. Therefore a careful selection of exposure conditions is necessary to achieve sufficient data quality. This comprises X-ray parameters and filters as well as the condition of the soil. The parameters were defined as such: The X-ray source was FXE-225.45, accelerating voltage 200 kV, total emission 200 μA, the filter 1 mm Cu, detector Perkin Elmer, 1024 × 1024 pixel 200 μm, scan period 50 minutes and a resolution of 141 μm. After segmentation voxel elements, which have a known volume and specific grey level for tubers, were used to determine tuber volume by calculating how many voxels are present in a reconstructed tuber image. The lower and upper threshold grey values were set at 2938 and 3963 after which mistakes were corrected for manually. Tuber volume calculations were performed using Image J software http://rsbweb.nih.gov/ij.