Genome-wide analysis of starch metabolism genes in potato (Solanum tuberosum L.)

Genome-wide analysis of starch genes in potato reveals novel isoforms

Enzymes involved in starch metabolism often belong to gene families encoding several isoenzymes. To our knowledge, this work presents the first genome-wide analysis of starch genes in potato. A comprehensive BLAST search strategy complemented by motif discovery and comparison to known sequences from Arabidopsis was applied aiming at the identification of all “starch gene” loci in potato. We found 77 loci coding for starch metabolism related enzymes belonging to different enzyme classes. Higher plants possess five gene classes encoding starch synthases, designated GBSS and SS1-4 [4]. In rice, two forms of GBSS were identified and eight genes encoding the four SS classes [41]. In the potato genome, we confirmed that GBSS which has been reported previously to be responsible for amylose biosynthesis in the starch granule, is encoded by a single gene and is expressed higher in tubers than in leaves [42]. A second transcript (DMT400003356) annotated as GBSS2 in the PGSC database had previously been described to possess soluble SS activity [43, 44] and was found to be the closest potato homolog to Arabidopsis SS2 and was therefore consequently designated as SS2. Moreover, it was shown that SS2 plays only a minor role in starch biosynthesis in tubers [43] which is in accordance with our expression analysis showing only a slightly higher expression in tubers compared to leaves although being upregulated during tuber development. In total, seven starch synthases were found in the potato genome (GBSS, SS1-6) most of which have been described in earlier studies [42–48]. However, no studies have analyzed the roles of SS5 and SS6 in potato yet, but our gene expression data suggest a possible role for SS5 in potato tuber starch biosynthesis. In contrast, SS6 is expressed to similar levels in leaves and tubers (Additional file 4) and its expression was not found to change significantly during tuber development (Fig. 5). Thus, further analyses are necessary to investigate the function of these genes during starch biosynthesis in potato. A recent publication described the phylogenetic relationship of SS from different plant species, including potato, confirming the presence of a fifth class of SS [49]. In addition the authors found maize SS5 to be highly expressed during the grain filling stage suggesting a role in starch biosynthesis [49], which is in agreement with our assumption.

In this study, enzymes were designated regarding to their annotation in Arabidopsis. In most cases, this was in accordance with isoform numeration of already described enzymes of potato. One exception concerns the numeration of isoforms within the SBE class where we identified four isoenzymes. Two of them share a very high sequence similarity to each other and have been denoted as SBE1.1 and SBE1.2 due to their homology to Arabidopsis SBE1. The deduced transcript sequences of these two genes, however, do not correspond to the previously published potato SBE1 sequences [46, 50, 51]. The gene product designated SBE1 in the aforementioned studies corresponds to SBE3 in this study. It was described as the major SBE isoform in potato tubers and was found to play a role in starch granule formation [51, 52]. This is in accordance with the expression profile during tuber development and tissue preference discovered in this study. Until now, only variants of two isoforms, SBE3 and SBE2, have been shown to act as branching enzymes in the amyloplast [46, 52]. The role of the two potato SBE1 paralogs identified in this study remains unclear. In Arabidopsis, SBE1 has an effect on embryogenesis and is essential for plant growth and development [53]. A direct implication of AtSBE1 in starch metabolism is unacquainted.

Comparative microarray analysis revealed tissue-specific gene expression

To identify tuber- and leaf-specifically expressed starch genes different microarray data sets were analyzed. To enable the analysis, specific microarray probes had to be assigned to the different starch genes and their respective isoforms. In general, our findings were in agreement with previously published gene expression analyses and showed a high reproducibility between the two microarray platforms. Tissue-specific expression of enzyme isoforms was for example found for PHO1a and PHO1b. PHO1b appeared to be preferentially expressed in leaves, while PHO1a was expressed higher in tubers, which is in accordance with previous findings [54, 55]. In the case of AGPase, most subunits are expressed slightly higher in tubers than in leaves according to our results. However, one isoform, namely APL1, was clearly expressed higher in leaves than in tubers. This is in contrast to findings from La Cognata et al. [56] who described tuber-specific expression of APL1 (designated AGP S3 in their work). The reliability of our results was confirmed by RNA-Seq data and by qRT-PCR using leaf and tuber samples. Genes showing tuber-specific expression were SuSy4, SBE3, SS5, GPT2.1 and SEX4. In contrast to the other tuber-specific isoforms, SEX4-specific transcripts were not up-regulated during tuber development which is consistent with the proposed role of the enzyme in starch degradation [57]. The activity of the main SuSy isoform in tubers, SuSy4, is connected to the onset of tuberization [58–60] and correlates well with transcript and tuber starch accumulation in potato [61]. Accordingly, SuSy4 overexpression led to an increased starch content and higher tuber yield in potato plants [62] supporting its suggested key role in starch metabolism. Similarly, SBE3 and GPT2.1 expression have been linked to tuber development and the accompanying accumulation of starch [63–65]. In this context, overexpression of GPT2.1 together with NTT resulted in increased tuber starch content and yield [66] indicating that expression and activity of GPT2.1 are closely related. The similarity between the expression patterns of these enzymes strongly supports the assumption of a coordinated transcriptional regulation of genes within the same pathway [63]. Moreover, these examples confirm that enzymatic activity of SuSy and GPT2.1 nicely correlates with transcript accumulation and that accumulation of starch metabolic enzymes is controlled at the transcriptional level. However, in other species activity of starch metabolic enzymes was shown to be additionally regulated by post-translational mechanisms. For example, phosphorylation of SuSy isoforms was shown to influence sub-cellular localization and protein stability [67]. Activity of SBE isoforms was reported to be regulated by protein phosphorylation and redox state [68].

Co-expression analysis reveals candidate regulators of starch biosynthesis

Co-expression analysis has previously been described to be a suitable tool for the identification of co-regulated genes [69, 70]. Assuming that proteins with regulatory functions have to be expressed at the same time or shortly before their target genes, the identification of candidate regulators should be possible by co-expression analysis. The great potential of this strategy has already been demonstrated in several studies including different plant species and tissues [15, 71–73]. One example is the identification of Rice Starch Regulator 1 (RSR1) by Fu and Xue [15] in a co-expression analysis similar to the approach used in this study. RSR1 was found to be negatively co-expressed with rice starch synthesis genes and was experimentally verified as a modulator of starch gene expression.

In this work, genes that were identified as being tuber-specifically expressed and exhibiting an expression pattern that coincides with starch biosynthesis in the potato tuber were used to search for potential transcriptional regulators, since they are so far not known. The number of genes identified to be co-expressed with SuSy4, GPT2.1 and SBE3differed between the two microarray platforms, and was about 10 times higher in the 8x60k experiments than in those performed with the POCI platform. One reason for this might be the sample selection of the 8x60k platform which basically consists of tuber samples in similar developmental stages while most samples taken from the 4x44k format were originally designed to reflect starch biosynthesis during tuber formation. Therefore, expression profiles derived from experiments using the POCI array were expected to be more specific with respect to the identification of putative regulators of starch biosynthesis in potato tubers. Moreover, we reasoned that co-expression of a regulator with its target genes should occur in all situations. Thus, candidate selection was made after comparing the results of the co-expression analyses of the three query genes in three different platforms each with many individual samples. Eventually we identified 40 genes that are consistently co-regulated with SuSy4, GPT2.1 and SBE3. Inspection of co-expressed genes revealed a strong over-representation of genes involved in primary carbon metabolism and development as well as genes encoding storage proteins. Tuber development and storage metabolism are known to be highly associated processes [39] which strengthens the significance of the retrieved candidates. Beside this, putative TFs co-expressed with the selected starch genes could be identified. They belong to different classes and none of them has been characterized in potato so far. Clearly, there is a strong enrichment of TFs associated with developmental processes and organogenesis like BOP2, LOB, PTL and LRP.

For PTL, a co-expression with SuSy4 and GPT2.1 in samples representing different tuber developmental stages could not be confirmed via qRT-PCR and PTL might therefore not be a good candidate for further analysis. The expression profiles obtained by qRT-PCR of the other three TF were in accordance with those of the microarray analysis (Fig. 7). Slight variations between qRT-PCR and microarray were found when comparing expression levels of SuSy4 or LOB on “Stage 5” and “grow” from the microarray to “Swollen stolon” and “growing tuber” samples used for qRT-PCR. In the microarray, highest gene expression was seen on “Stage 5”, while in the qRT-PCR expression peaked in growing tubers (Fig. 7). Nevertheless, an increasing expression level was always associated with tuber formation. A possible explanation for this disagreement might be slightly different developmental stages of the samples used for the analyses. For the microarray defined stages of tuber development were sampled [20, 39], while for the qRT-PCR swollen stolons of different developmental stages were pooled. Furthermore, the growing tubers for the microarray experiment were monitored by X-ray CT analysis to determine their growth velocity, while the tubers sampled for qRT-PCR were taken from plants during their growth period, assuming that the tubers were still growing. Despite these small differences between different experiments, expression levels of LOB, TIFY5a and WRKY4 correlate well with SuSy4 and GPT2.1 (Table 2). Thus they might be interesting candidates for further analyses.

In Arabidopsis, BOP2 and its close homolog BOP1 regulate the expression of LOB-domain containing proteins [74]. LOB expression has been found in the boundary regions between meristematic tissue and developing lateral organ primordia of the shoot apical meristem and the roots [75]. A similar spatial expression is exhibited by LRP1 of Arabidopsis which has been shown to be expressed in root primordia in early developmental stages [76]. In maize the localization of LRP in lateral root primordia was confirmed and it was demonstrated that LRP expression was auxin-inducible [77]. A link to auxin-signaling may also be established by the closest homolog of potato WRKY4 in Arabidopsis. Based on sequence similarity, the closest homolog in Arabidopsis is WRKY23 which has been linked to auxin-signaling in root development [78, 79]. A role of auxin in tuber initiation has been suggested [80] but a direct link to starch biosynthesis is missing. The expression patterns of these TFs suggest that there are interesting candidate genes which may directly or indirectly control starch biosynthesis and that more detailed investigation of their role is worthwhile.