We conducted a targeted analysis of tuber secondary metabolism in potatoes grown in a modern agricultural system that maximizes yield and essentially ameliorates the extreme environmental stresses a wild plant would naturally encounter. To better understand the metabolism of stress inducible compounds when stresses are deliberately minimized, secondary metabolites were evaluated in potatoes grown in multiple locations that encompass the general range of environmental conditions a U.S. crop would encounter. Environmental variables among locations included marked differences in day and night temperatures, light intensity, day length, cloud cover, humidity, soil type, time of planting and coastal versus inland locations. The secondary metabolites assessed are of particular interest because, in addition to their physiological roles, they are major determinants of the nutritional value of potatoes. Results showed that environmental influences occurring under managed conditions are sufficient to significantly impact the nutritional value of potatoes, with a two-fold or greater change seen for some phytonutrients. Importantly, over two-fold differences were seen in some metabolites between the Texas and Florida locations, showing significant environmental variation can occur in typical growing locations. Such substantial changes would be predicted to be especially relevant for staple crops because the high consumption means even modest alteration in phytonutrients will have a dietary impact.
We suspected the Alaskan potatoes would have higher amounts of some secondary metabolites, given the effects light and temperatures are known to have on plant phenylpropanoid and carotenoid pathways. The Wiseman location is north of the Arctic Circle and does not experience darkness in mid-summer. Total phenolics were highest in Wiseman and generally higher in the Alaskan samples than the southern samples. The spread from low to high was a somewhat surprisingly modest 28% as measured by the FC method. However, when individual phenylpropanoids were measured by HPLC substantial differences were found, including a two-fold difference in CGA. This is significant because CGA is typically most abundant phytonutrient in tubers and has been reported to constitute about 90% of tuber phenolic content in some genotypes [38
]. Stimuli known to increase CGA biosynthesis in potatoes include light [4
], while seasonal variation in CGA content occurs in evergreens [39
]. Interestingly, the lower CGA content of the southernmost samples was partly offset by the several-fold higher amounts of 3CGA and 4CGA (Figure ).
expression did not correlate with CGA concentrations or other phenylpropanoid pools (Figure , and Figure ). Curiously, a better correlation was observed between all of the other phenylpropanoid genes and CGA, than between HQT
and CGA, as calculated using Pearson product moment correlation coefficients (Figure ). Gene expression analysis was conducted on mature tuber tissue sampled at the end of the season and therefore would not detect transient changes in expression occurring earlier in development, prior to the sampling point. However, these analyses do allow comparison of gene expression and metabolite levels in mature tubers at the same time point. CGA is synthesized by HQT
, although additional pathways are possible [34
]. Promoter analysis identified two elements (EIRE and LTR) that were present in HQT
, but not the other phenylpropanoid genes. Whether these contribute to the difference in HQT
expression relative to the other phenylpropanoid genes is unclear. It has not yet been definitively shown that HQT
is the only or even major route to CGA synthesis in tubers, although based on results with tomato this seems probable [35
]. A weak correlation between HQT
mRNA expression and CGA concentrations was previously reported in coffee grains, artichoke leaves and drought stressed potatoes [27
The strongest correlation between core phenylpropanoid pathway gene expression and metabolite pools occurred with PAL
, emphasizing the importance of this gene in modulating tuber phenylpropanoid metabolism (Figure ). The lower phenylalanine pools in samples with higher PAL gene and enzyme activity is consistent with more flux through the pathway. Similarly, reduced PAL activity and higher levels of phenylalanine in the southern samples may reflect less carbon flow into the phenylpropanoid pathway and slower turnover of phenylalanine. The strong correlation between UFGT
expression and the more abundant tuber hydroxycinnamic acids like CGA and FQA2 suggests that increased anthocyanin biosynthesis may lead to increased amounts of upstream phenylpropanoids outside of the anthocyanin branch. This could reflect coordinate up-regulation of the overall phenylpropanoid pathway in response to increased anthocyanin biosynthesis. The anthocyanins would act as a carbon sink and increase carbon flow from the shikimate pathway into the phenylpropanoid pathway. Such a mechanism could explain why anthocyanin-rich potatoes also tend to have greater amounts of colorless phenolics like CGA [29
]. An interesting finding was the inverse relationship seen between some isomers or among closely related compounds in which the more abundant isomers or family member were positively correlated with phenylpropanoid transcript abundance, but the less abundant ones were negatively correlated. This was observed among the chlorogenic acids, ferulic acids and anthocyanins and shows that tubers have the ability to increase or decrease individual phenylpropanoids even though the pathway is coordinately regulated. This may also allow tubers to partially, but not completely, compensate for a decrease in certain compounds. For example, the Alaskan samples had higher amounts of CGA, but the southern tubers higher amounts of 3- and 4CGA (Figure ) and CGA pools inversely correlated with nCGA and cCGA, which were strongly correlated with each other.
The most abundant anthocyanins strongly correlated with PAL
expression and with each other (Figure ) and were significantly higher in Alaskan samples (Figure and Figure ) with over a two-fold difference seen in petunidin-3-coum-rut-5-glu (PtCRG), the most abundant anthocyanin. However, expression of individual anthocyanins did not necessarily parallel total anthocyanin levels. The Wiseman samples had the most PtCRG and total anthocyanins, but samples from the Southern U.S. had significantly higher amounts of two less abundant anthocyanins, PtD and MaCRG (Figure ). This suggests that individual anthocyanins are differentially synthesized in response to environmental signals. These results are consistent with a report of higher anthocyanins in potatoes grown in Colorado than Texas [25
]. Anthocyanin content was also found to differ significantly in Bog Bilberries grown in north or south Finland [13
Anthocyanins are regulated at transcriptional and translational levels, and are influenced by light, temperature and biotic stresses [2
], and accumulation is favored under lower temperatures [45
]. Anthocyanins can decrease under higher temperatures [42
] and the mechanism involves both anthocyanin degradation and a decrease in transcription [46
]. Consistent with transcriptional control of tuber anthocyanins, samples from Texas and Florida had lower expression of the key flavonoid genes CHS, DFR
that correlated with anthocyanin content (Figure and Figure ).
Sucrose, but not glucose concentrations, positively correlated with the most abundant phenolic acids and anthocyanins and with phenylpropanoid gene expression except for HQT
, but did not correlate with carotenoids or early polyamine pathway genes. Sucrose had an inverse correlation with the shikimate derived amino acids (Figure ), which is consistent with some previous reports [47
]. These amino acids were positively correlated with each other, suggesting coordinate regulation.
A surprising finding was how carotenoid expression varied among locations. Carotenoid composition rather than the total carotenoid content showed the greatest differences (Figure and Figure ). The carotenoid profiles seen in HPLC chromatograms varied enough to appear to be from different genotypes, with only trace amounts of some compounds present in one location but abundant in another (Additional file 3
). For instance, in the Alaskan samples violaxanthin was the predominant carotenoid, but ten-fold less was present in Florida tubers, which instead had higher amounts of zeaxanthin, which in turn was only present in trace amounts in all the other locations (Figure ). This might be due to the xanthophyll cycle in which zeaxanthin and violaxanthin are intercoverted to protect photosystem II from excessive light energy and higher temperatures [49
]. The Florida potatoes also had higher amounts of antheraxanthin, an intermediate in the violaxanthin/zeaxanthin reversible interconversion. High light intensity or temperature leads to increased formation of zeaxanthin by violaxanthin de-epoxidase (VDE), whereas low light stimulates the violaxanthin formation by zeaxanthin epoxidase [50
]. However, neither ZEP
was more highly expressed in the Florida potatoes than other samples, suggesting the differences were not due to transcriptional control of these two genes. Ascorbate has been shown to stimulate the de-epoxidation of violaxanthin to zeaxanthin [51
] and the Florida samples had greater amounts of ascorbate (Figure ). This might favor zeaxanthin formation, but Fairbanks tubers had similar amounts of ascorbate yet only trace amounts of zeaxanthin.
The Texas potatoes were unique in that they had a more than double the amount of lutein, indicating that the α-carotene branch of the pathway was more active. A branch point in the carotenoid pathway occurs at lycopene, which is the substrate for two competing enzymes, LYC-e and LYC-b (Figure ). Among the carotenoid genes examined, CHY2, LCYe
showed strong correlations (Additional file 4
). LCY-e transcript was more highly expressed in the Texas potatoes and catalyzes the ε cyclization of lycopene. The LUT1
product catalyzes the terminal step in the pathway, but LUT1
was not more highly expressed in the high lutein tubers. A poor correlation between LUT1
expression and lutein content was seen in potatoes silenced for LCY-e
] and strong correlations have generally not been reported between transcript levels of genes downstream of lycopene and individual carotenoids [54
]. The Texas potatoes did not have higher concentrations of any carotenoid other than lutein, yet they showed the highest expression of ZEP
in addition to LCY-e
. A recent study suggested increased carotenoids in potatoes is due to increased metabolic flux and found correlations between carotenoid genes and metabolites [56
], which suggests transcript abundance does influence carotenoid concentrations among cultivars and differs from our study that examined expression within the same genotype in different conditions.
An interesting question is to what extent the tuber directly perceives environmental cues that alter carotenoid or phenylpropanoid composition versus the foliage modulating tuber expression. Several phenylpropanoid gene promoters are known to have light regulatory elements such as Box 4, G box, I box, and L box [57
]. These elements were also present in most potato phenylpropanoid gene promoters but their role in tubers not exposed to light is unclear.
The observed changes in potato phytonutrients show a potential to customize the nutritional value of a potato crop by choice of cultivar and environment, which may become increasingly relevant with advances in nutrigenomics. For example, the markedly higher anthocyanin amounts in the Alaskan potatoes provide a rationale to take advantage of local environmental conditions by growing red- or purple-flesh potatoes. White-flesh potatoes grown in Alaska would not provide significant anthocyanins, which also illustrates that there are genotypical components to the response to environmental signals. Similarly, the higher amounts of lutein in the Texas potatoes and zeaxanthin in the Florida potatoes raise questions about the extent to which growth can be manipulated to give a desired phytonutrient profile.