Diurnal regulation of cyanogenic glucoside biosynthesis and endogenous turnover in cassava

Abstract Cyanogenic glucosides are present in many plants, including eudicots, monocots, and ferns and function as defence compounds based on their ability to release hydrogen cyanide. In this study, the diurnal rhythm of cyanogenic glucoside content and of transcripts and enzymes involved in their biosynthesis was monitored in cassava plants grown in a glasshouse under natural light conditions. Transcripts of CYP79D1, CYP79D2, CYP71E7/11, and UGT85K5 were at minimal levels around 9 p.m., increased during the night and decreased following onset of early morning light. Transcripts of UGT85K4 and HNL10 showed more subtle variations with a maximum reached in the afternoon. Western blots showed that the protein levels of CYP71E7/11 and UGT85K4/5 decreased during the light period to a near absence around 4 p.m. and then recovered during the dark period. Transcript and protein levels of linamarase were stable throughout the 24‐hr cycle. The linamarin content increased during the dark period. In the light period, spikes in the incoming solar radiation were found to result in concomitantly reduced linamarin levels. In silico studies of the promoter regions of the biosynthetic genes revealed a high frequency of light, abiotic stress, and development‐related transcription factor binding motifs. The synthesis and endogenous turnover of linamarin are controlled both at the transcript and protein levels. The observed endogenous turnover of linamarin in the light period may offer a source of reduced nitrogen to balance photosynthetic carbon fixation. The rapid decrease in linamarin content following light spikes suggests an additional function of linamarin as a ROS scavenger.

In cassava, the first step of the pathway for biosynthesis of linamarin and lotaustralin ( Figure 1) is catalyzed by the cytochrome P450 enzymes CYP79D1 and CYP79D2, which both convert valine and isoleucine to the corresponding oximes (Andersen, Busk, Svendsen, & Møller, 2000). The conversion of the two oximes formed into a-hydroxynitriles is catalyzed by either CYP71E7 or CYP71E11 (Jørgensen et al., 2010). The hydroxynitriles are then glucosylated by either UGT85K4 or UGT85K5 to produce linamarin and lotaustralin .
Using embryogenic callus cultures, efforts have been undertaken to obtain acyanogenic cassava by RNA interference technology with CYP79D1 and CYP79D2 as targets. Transformed plants, in which the cyanogenic glucoside content was reduced to <25% of the cyanide potential of wild-type plants, had long and slender stems with long internodes and typically did not develop roots.
When the nitrogen supply was increased, growth was partly restored (Jørgensen et al., 2005). This suggests that the physiological role of cyanogenic glucosides extends beyond their role as defence compounds (Møller, 2010). The availability of nitrogen and carbon dioxide as well as drought stress affect the cyanogenic glucoside content and growth of cassava (Gleadow, Evans, McCaffery, & Cavagnaro, 2009;Gleadow & Moller, 2014;Rosenthal et al., 2012). The cyanogenic glucoside dhurrin is synthesized and accumulates in the developing sorghum (Sorghum bicolor) seed but is turned over in the course of seed maturation so that the mature sorghum seed does not contain dhurrin (Nielsen et al. 2016). When sorghum seeds germinate, dhurrin is rapidly synthesized in the young seedlings reaching levels of 30% of the dry mass in the tip of etiolated seedlings (Busk & Møller, 2002;Halkier & Møller, 1989). The acyanogenic sorghum line tcd1, obtained by classical mutagenesis, exhibits slow germination and reduced height of adult plants compared to wild-type plants (Blomstedt et al., 2012). In the rubber tree (Hevea brasiliensis), the cyanogenic glucoside content is influenced by time of day, light and by latex tapping (Kongsawadworakul et al., 2009).

Statement of significance
The biosynthesis and endogenous turnover of cyanogenic glucosides in cassava show strong diurnal regulation at the transcript and protein levels. In addition to their classical function as plant defence compounds, the study points to a role of cyanogenic glucosides as mobilizable storage compounds of reduced nitrogen and carbon and as ROS scavengers.
function of cyanogenic glucosides as storage forms of reduced nitrogen and carbon (Nielsen et al., 2016;Picmanova et al., 2015).
In this study, we demonstrate that the production and accumulation of cyanogenic glucosides follow a strong diurnal rhythm with turnover of a significant fraction of the cyanogenic glucosides in the light and resynthesis during the night. The synthesis of linamarin is controlled both at the transcript and protein levels.

| RESULTS
To identify factors possibly involved in regulating biosynthesis and endogenous turnover of cyanogenic glucosides at the molecular level, an initial in silico analysis of the promoter regions of the genes encoding enzymes involved in cyanogenic glucoside biosynthesis and hydrolytic bioactivation was carried out. Based on the findings from the in silico study, an experiment investigating the diurnal variation of cyanogenic glucoside metabolism was set up.  were found in all promoters with most found in pCYP79D1, pUGT85K5, and the promoters of the hydrolytic genes. pCYP79D2 showed very few motifs compared to pCYP79D1, thus indicating different roles for the two homologs with regard to circadian regulation. In the light-related motifs category, pCYP79D1 and the two pUGTs contained the highest number of motifs. This combined with the circadian category indicated that the regulation of pCYP79D1 and the pUGTs is influenced by light. pCYP79D2, the two pUGTs, and pHNL10 had similar or higher number of motifs related to nutrient-related regulation than the rest of the analyzed promoters. pCYP79D1 contained in general more motifs than its pCYP79D2 homolog. Whereas pCYP71E11 exhibited a uniquely high number of abiotic stress-related motifs, pCYP71E7 contained a high number of biotic stress-related motifs. For comparisons, the promoters of Rubisco small subunit (pRubSS) and LOX2 (pLOX2) were analyzed as representative of genes encoding enzymes related to photosynthesis and biotic stress, respectively (Jung et al., 2007;Spoel et al., 2003). pRubSS contained a small number

| Cyanogenic glucoside content
The stoichiometric ratio between the two cyanogenic glucosides linamarin and lotaustralin present in cassava is general around 90:10 (Lykkesfeldt & Møller, 1994;Nartey, 1968). In this study, only data for linamarin are presented, as the calculated ratio was consistent and close to the general ratio. An ANOVA test was performed for the data set and showed p < .001, F > F crit .
LC-MS analysis demonstrated that the linamarin content varied during the day and night cycle with a difference of 35% between the highest and the lowest levels ( Figure 3). From 9 a.m. until noon, the linamarin content was found to decrease moderately. The content

| Protein levels of enzymes catalyzing cyanogenic glucoside biosynthesis and hydrolytic bioactivation
The protein levels of CYP71E7/11 and UGT85K4/5 displayed oscillatory dynamics throughout the entire light period as determined by immunoblot analysis (Figure 4). The antibodies used do not distinguish between the two homologs of each protein. The levels of the CYP71E7/11 and UGT85K4/5 proteins were at their highest around 9 a.m. and decreased to almost nondetectable levels between 3 p.m. and 5 p.m. This synchronization of protein expression pattern and/or degradation indicated that the proteins are coregulated by a yet unknown factor or factors. Several immune responsive protein bands were observed using the UGT85 antibody. When the antibody was tested against crude protein extracts of the first fully unfolded leaf of cassava plants grown under at lower light intensities, a single protein band was observed . We therefore conclude that the additional bands observed in this study represent proteolytic degradation products of the UGT85K4/5 proteins. In contrast to the CYP71E7/11 and UGT85K4/5 proteins, linamarase and HNL did not show any distinct variation in their levels as com-

| DISCUSSION
In this study, the effect of the diurnal rhythm on the metabolism and content of the cyanogenic glucoside linamarin in cassava was investigated using plants grown in a glasshouse.
The content of cyanogenic glucosides has previously been reported to vary between vegetative cassava clones propagated from stem cuttings, even when grown under identical and sterile conditions in a climate chamber (Jørgensen et al., 2005). As for other bioactive natural products, the accumulation of cyanogenic glucosides is controlled by their rate of biosynthesis relative to the rate of hydrolytic bioactivation and endogenous turnover. Deciphering how these processes are controlled by variations of transcript and protein levels and by post-translational modifications is highly complex. Unfortunately, the enzymes and genes involved in the endogenous turnover pathway of cyanogenic glucosides are yet to be identified. The endogenous turnover products of linamarin are linamarin amide, linamarin acid, and linamarin anitrile, and these may all be detected in cassava leaves but in levels representing around 0.1% of that of linamarin (Nielsen et al., 2016;Picmanova et al., 2015). The low amounts of endogenous turnover products most likely reflect their rapid further metabolism and incorporation into primary metabolism.
The results demonstrated that the levels of the biosynthetic enzymes CYP71E7/11 and UGT85K4/5 decreased upon the onset of morning light with minimum levels between 4 p.m. and 6 p.m. A similar decrease was observed for the transcripts of CYP79D1, CYP79D2, CYP71E7/11, and UGT85K5. The protein and transcript levels for these increased during the dark period. A similar recovery of the linamarin levels was observed in the dark period, whereas the pattern in the light was rather complex. Despite rapid changes in protein and transcript levels, only moderate changes were observed on metabolite level. The changes in linamarin levels related to the diurnal rhythms were overlaid by decreases caused by abrupt increases in light intensity caused by changes from a cloudy to a clear sky (Figure 3). Prior to this experiment, the experiment was repeated twice in minor scale, and the same trends were observed in all three experiments (Figure 3, Figures S1 and S2). Similar results have been reported in green leaves of Olinia ventosa, where exposure to strong light resulted in the conversion of prunasin into its corresponding amide (Sendker & Nahrstedt, 2009). The amide formation observed under high light irradiation may not necessarily reflect the operation of the endogenous turnover pathway. It is most likely assigned to the ability of cyanogenic glucosides to sequester reactive oxygen species (ROS) in a nonenzymatic Radziszweski process by which the nitrile functional group is converted into an amide group (Møller, 2010;Sendker & Nahrstedt, 2009). In rubber tree (Hevea braziliensis), more than twofold decrease in linamarin content was observed during the light period, and the linamarin content of shade leaves was higher than in leaves exposed to direct light (Kongsawadworakul et al., 2009). In developing sorghum seedlings, the rise in the content of the cyanogenic glucoside dhurrin was much higher during the night compared to the light period (Adewusi, 1990  The classical function of cyanogenic glucosides is their role as phytoanticipins in a two-component insect defence system which by the action of b-glucosidases upon cell disruption results in the detonation of a hydrogen cyanide bomb (Conn, 1969;Gleadow & Moller, 2014;Morant et al., 2008

| Metabolite analysis
A single leaf disk (⌀ = 10 mm) was excised from the center part of each of ten leaves, excluding the midrib. To avoid loss of labile metabolites, each single leaf disk (10 biological replicates) was extracted in boiling 85% MeOH (300 ll, 3 min, 1.5 ml sealed screwcap tube) immediately after harvest. After boiling, the tubes were chilled on ice and the liquid fraction transferred to a glass vial and SCHMIDT ET AL.
| 7 stored at 5°C and analyzed within 4 days. Aliquots (15 ll) were diluted with 135 ll water and filtered (0.45 lm 96-well filter plate, 2,000 g, 10 min) and hereafter transferred to a new glass vial for LC-MS and analyzed as in Jørgensen et al. (2005). The data were analyzed using DataAnalysis 4.0 (Bruker Daltonics GmbH), and statistical analysis was performed in Excel 2010 (Microsoft).
After electrophoretic separation, the proteins were transferred to