Cyanogenic glucosides were analyzed as described previously (Pičmanová et al., 2015). Samples (100 mg) were ground to a fine powder in liquid nitrogen, mixed with 400 μL 85% methanol, boiled 5 min, placed on ice and centrifuged (5 min, 20,000 × g). Aliquots (20 μL) of the supernatant were mixed with 70 μL of water and 10 μL of 500 μM internal standard (linamarin) and filtered through a filter plate (0.45 μm, Millipore) by centrifugation (5 min, 1,107 × g).

LC–MS/MS was carried out using an Agilent 1100 Series LC (Agilent Technologies) coupled to a Bruker HCT-Ultra ion trap mass spectrometer (Bruker Daltonics). A Zorbax SB-C18 column (Agilent; 1.8 μm, 2.1 mm × 50 mm) maintained at 35°C was used for separation. The mobile phases were: (A) water with 0.1% (v/v) HCOOH and 50 mM NaCl; (B) acetonitrile with 0.1% (v/v) HCOOH. The gradient program was: 0–0.5 min, isocratic 2% B; 0.5–7.5 min, linear gradient 2–40% B; 7.5–8.5 min, linear gradient 40–90% B; 8.5–11.5 min isocratic 90% B; 11.6–17 min, isocratic 2% B. The flow rate was 0.2 ml⋅min^-1 but increased to 0.3 ml⋅min^-1 in the interval 11.2–13.5 min. ESI–MS² was run in positive mode. The data was analyzed using the Bruker Daltonics programme Data Analysis 4.0. Extracted ion chromatograms for specific [M+Na]⁺ adduct ions (as NaCl is added to one of the mobile phases, the great majority of adducts formed are [M+Na]⁺; we could also see [M+H]⁺ and [M+NH₄]⁺, but these are minute in comparison with the sodium adducts) and their MS² profiles were used to identify the compounds.

Table 3 shows the names, structures, and retention times of all the compounds detected in this study. Amygdalin was bought from Sigma–Aldrich. Prunasin was chemically synthesized (Møller et al., 2016). Prunasin amide, prunasin acid, prunasin anitrile, 1-O-benzoyl-β-D-glucopyranose, prunasin-6′-β-D-apioside and prunasin-6′-β-D-xyloside were chemically synthesized (Motawia MS, unpublished work). The reference compounds were used for absolute quantification in a range of concentrations from 0.5 to 125 μM. As for the relative quantifications presented, the ionization efficiency of prunasin and its derivatives may differ by a factor of approximately two, and hence the ratios expressed as percentages of prunasin content are correct within this span (Pičmanová et al., 2015). The MS and MS² spectra observed for each compound were in agreement with the spectra previously reported (Pičmanová et al., 2015).

Samples were assayed in two to three technical replicates, except for the last time point of the prunasin content in S3067.

Quantitative real-time polymerase chain reaction (qRT-PCR) based expression analysis was carried out on 12 selected genes using three reference genes (TEF2, 18s rRNA, and RPL13) (Table 4). The targeted gene sequences were based on homologous genes derived from different Prunus species and preliminary transcriptomic data from sweet cherry ‘Burlat’ flower buds (Ionescu et al., 2017).

Sweet cherry ‘Burlat’ flower buds samples were obtained from branches kept for 1, 3, 7, 10, 15, and 17 days at controlled conditions. Frozen plant material was ground with mortar and pestle in liquid nitrogen. For each sample, total RNA was extracted using the Spectrum^TM Plant Total RNA Kit (Sigma–Aldrich, St. Louis, MO, USA) and 500 ng of RNA was used to generate cDNA using the iScript^TM cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA). Gene-specific primer pairs were designed for target and reference genes using two web based tools: NCBI’s Primer-BLAST¹ and IDT’s² PrimerQuest^© (Table 4). Primer efficiencies were 82 ± 12% and their sequence specificity was determined by sequencing the amplicon and comparing it to the original coding sequence used for initial primer design (see Supplementary Data Sheet 1, DS1). Obtained sequences were aligned to the associated coding sequences using a local alignment with Needleman–Wunsch algorithm (Needleman and Wunsch, 1970). Herein, sequence coverage was 75 ± 6% and identity was 83 ± 9%.

qRT-PCR was performed using a CFX384^TM real-time PCR detection system. Reactions were conducted in 8 μl volume using the DyNAmo Flash SYBR Green qPCR Kit (Thermo Fisher Scientific, Waltham, MA, USA) with each reaction containing 1x DyNAmo Flash SYBR Green qPCR Mix (2x), 5 ng of cDNA template and 625 nM of both forward and reverse primer. The following PCR protocol was used: 7 min at 95°C, [10 s at 95°C, 30 s at 60°C, 1x plate read] × 40 cycles, 1 min at 60°C. A melting curve was performed for each reaction. Further, no template controls as well as no RT controls were included. A standard curve for TEF2 was used as interrun control using the deduced PCR efficiency as factor for interrun deviation. Relative gene expression levels were computed from the qPCR data using the ΔΔCq calculation method (Livak and Schmittgen, 2001). Therein a normalization factor based on the expressional variation of three reference genes among the examined samples was used. This factor was obtained using geNorm version 3.5 (Vandesompele et al., 2002).

The CNglc prunasin (Table 3) was detected in all five almond cultivars during the entire developmental period of the buds from dormancy to flowering (Figure 3A). Prunasin was also detected under controlled conditions in flower buds of the sweet cherry ‘Burlat,’ but in levels approxiamtely 10-fold lower than in almond (Figure 4A). Amygdalin, the other CNglc present in almond, was detected in all five cultivars in minute amounts, approximately 200-fold lower, compared to prunasin (Figure 3E). This is within the range (37–300-fold lower) that had been previously observed in two almond cultivars (Ramillete-sweet and S3067-bitter), when prunasin and amygdalin were measured in the leaves of almond trees after the almonds had been harvested (Figure 5 in Sánchez-Pérez et al., 2008). In sweet cherry flower buds, no amygdalin was detected. The di-glucoside amygdalin is present in very minute amounts compared to the monoglucoside prunasin and this is in agreement with a previous observation (Frehner et al., 1990; Dicenta et al., 2002; Sánchez-Pérez et al., 2008). The situation is reverse in bitter almond seeds where amygdalin is the dominating cyanogenic glucoside. In vegetative parts of the tree, prunasin is always the dominating cyanogenic glucoside present.

In general, during the entire dormancy-flowering period, the level of prunasin was highest in the early cultivars Achaak and Desmayo, followed by S3067. Lauranne and Penta contained the smallest amounts of prunasin (Figure 3A). The prunasin profiles obtained shared clear relations to the dates of dormancy breaking and flowering time. In all five almond cultivars as well as in the single sweet cherry cultivar, prunasin started to accumulate at the time of dormancy release or shortly thereafter and reached its maximum just before flowering took place. This may suggest that prunasin plays a role in flower development after dormancy is broken.

Dissection of almond flowers enabled detection of prunasin and minute amounts of amygdalin in pistils, petals and sepals of all five almond varieties (Supplementary Figure S1). S3067 was the only variety where prunasin could be detected in the pollen, but the amount of pollen available was too low to acquire biological and technical replicates. In relation to this, amygdalin content has previously been reported in almond pollen at about 1890 ppm (London-Shafir et al., 2003) and reported to deter inefficient pollinators, thus allowing more efficient pollination by honeybees, adapted to tolerate higher levels of amygdalin. Prunasin had previously been reported detected in sepals, petals, pistils, and pollen of flowers from bitter and sweet almond cultivars (Abarrategui, 2010). Amygdalin levels were almost zero, except in the bitter cultivars.

In the case of Lotus japonicus, the two aliphatic CNglcs linamarin and lotaustralin are present throughout in the flower tissue (Lai et al., 2015). As mentioned previously (Figure 1B), bioactivation of the CNglc takes place only when specific β-glucosidases come into contact with their corresponding substrate. In L. japonicus the reproductive organs are only cyanogenic when a specific β-glucosidase BGD3 is expressed (Lai et al., 2015). Hydrogen cyanide release was derived specifically from the keel and enclosed reproductive organs of the flower. Sepals, wings, buds, and pods also contained the cyanogenic glucosides linamarin and lotaustralin, but no release of any hydrogen cyanide from these tissues was observed because the β-glucosidases were not present in these tissues (Lai et al., 2015). It needs to be investigated whether or not a particular β-glucosidase might also be expressed in almond and cherry flower buds.

The presence of CNglcs in flowers of other species has previously been reported. Within the Prunus genus, prunasin was quantified in flowers of P. avium, whereas amygdalin was not detected (Nahrstedt, 1972). Prunasin as well as amygdalin were identified in flowers of P. yedoensis Matsum (Matsuoka et al., 2011). Five different CNglcs were also found in flower buds of Eucalyptus camphora subsp. humeana, namely prunasin and the diglucosides amygdalin and eucalyptosins A, B, and C (Neilson et al., 2011). In Turnera ulmifolia L., the content of CNglcs decreased to zero when the plant began to flower (Schappert and Shore, 2000) indicating complete endogenous turn-over of CNglcs for alternative uses. CNglcs have also been detected in flower tissues of Grevillea species, Linum usitatissimum L. (flax), L. japonicus L., Ryparosa kurrangii B.L. Webber (rainforest tree) and E. camphora L.A.S. Johnson and K.D. Hill (Lamont, 1993; Niedźwiedź-Siegieñ,, 1998; Forslund et al., 2004; Webber and Woodrow, 2008; Neilson et al., 2011).

In addition to prunasin and amygdalin, structurally related derivatives were also found in the flower buds of the five almond cultivars (Table 3 and Figures 3B–D, 5), in almond pistils, sepals and petals (Supplementary Figure S1) and in the one sweet cherry cultivar analyzed in this study (Table 3 and Figures 4B, 6). The prunasin derivatives prunasin amide, prunasin acid, prunasin anitrile, and the diglycoside prunasin pentosides were all present in amounts much lower than prunasin. In contrast, the non-cyanogenic diglycoside prunasin anitrile pentosides (prunasin anitrile arabinoside and xyloside in almond and most probable prunasin anitrile apioside in cherry) were highly abundant at certain stages of flower development (Figures 5, 6).

The content of prunasin amide (Figures 3B, 4B) displayed a very interesting and consistent pattern in the five almond and the single cherry cultivar analyzed. Prunasin amide was not detectable until it peaked very close to flowering time. In almond, the highest amount of prunasin amide was found in the earliest cultivar (Achaak). In all studied cultivars, the peak of prunasin amide coincided with a decrease in prunasin levels, indicating turnover of prunasin into its amide. The conversion of prunasin to prunasin amide may occur non-enzymatically via the Radziszewski reaction in the presence of hydrogen peroxide (Sendker et al., 2016). Hydrogen peroxide is produced during flower development (Kuroda et al., 2002). Although present in small amounts, formation of prunasin amide may thus serve as a quenching reaction to avoid toxic hydrogen peroxide levels (Møller, 2010). Alternatively, prunasin amide might be formed from prunasin catalyzed by a bifunctional nitrilase or by a nitrile hydratase (Pičmanová et al., 2015).

The presence of prunasin acid (Figure 3C) was detected at the beginning of almond flower bud development, although with relatively high standard error margins. In the mid-late cultivars S3067, Lauranne and Penta, small amounts of prunasin acid were observed to accumulate at the time point of flowering. Prunasin acid is likely formed from the prunasin amide (Figure 7). The levels of prunasin acid in the cultures Achaak and S3067 were close to zero. Low amounts of prunasin anitrile were accumulated in the almond cultivars, with peak levels before endodormancy release (Figure 3D).

In addition to the monoglucosides described above, two diglycosides (pentosides) derived from prunasin were identified in this study. Absolute quantification was not possible due to the lack of reference compounds. Therefore, we expressed the levels of these compounds as % of prunasin (Figures 5, 6). The levels of prunasin pentoside in almond (potentially a mixture of two prunasin pentosides) (Pičmanová et al., 2015) were higher at the beginning in the dormant stage, where CU had not yet accumulated (Figure 5A). All almond cultivars exhibited the presence of prunasin anitrile pentoside during endodormancy release, reaching relative amounts of up to 2000% of prunasin (e.g., Achaak, Figure 5B). In cherry, the levels of prunasin anitrile apioside increased toward the end of the experiment (Figure 6A).

In senescent leaves of P. laurocerasus L., novel benzoic acid esters have recently been reported as formed from prunasin (Sendker et al., 2016). This inspired us to investigate the possible presence of benzoic acid derivatives in almond and sweet cherry flower buds. A compound identified as β-D-glucose-1-benzoate was indeed found to be present in high amounts compared to prunasin in the flower buds of all studied almond cultivars as well as in the cherry cultivar (Figures 5C, 6B). β-D-Glucose-1-benzoate was suggested to be formed as a novel extension of the oxidative catabolism of prunasin (Sendker et al., 2016). The amount of accumulated β-D-glucose-1-benzoate is high compared to the prunasin level implying that β-D-glucose-1-benzoate might also be synthesized by a different route in the flower buds. Moreover, in almond and cherry flower buds, the formation of β-D-glucose-1-benzoate from the corresponding aldehyde could potentially be connected to the release of hydrogen peroxide during dormancy release. As mentioned previously, hydrogen peroxide has been implicated in flower development in Japanese pear (Pyrus pyrifolia Nakai) (Kuroda et al., 2002).

These results are in accordance with a recent study reporting the presence and structural identification of CNglc derived metabolites including di- and tri-glycosides in cassava, sorghum, and almond (Pičmanová et al., 2015). The amides, acids and anitriles derived from prunasin and amygdalin were identified in seedlings of the bitter almond cultivar S3067. The levels of the derivatives of prunasin and amygdalin were generally much lower than those of their mother compounds. Prunasin amide, acid and anitrile were found in low levels in seeds, roots, shoots, and leaves of the seedling and at different stages of germination. Prunasin acid was the most abundant derivative in seeds, shoots, and leaves and prunasin anitrile was most abundant in roots. An important increment of the prunasin derivatives was observed in the seed at the beginning of the germination (Pičmanová et al., 2015). Similarly, minor components related to CNglcs were detected in P. persica seeds: amygdalin acid, prunasin acid, benzyl gentiobioside and benzyl glucoside (Fukuda et al., 2003). The latter two compounds correspond to the amygdalin anitrile and prunasin anitrile compounds denoted in our study.

Our current study provides further evidence in support of the conclusions by Pičmanová et al. (2015) that CNglcs occur together with their putative structural derivatives: amides, acids and anitriles. In this respect, it was suggested that these derivatives could play a role in the recycling of reduced nitrogen. An alternative endogenous turnover pathway was proposed in which CNglcs are converted to non-CNglcs, without release of HCN (Figure 1D). Hypothetically, amides, acids, and anitriles are produced from CNglcs in this turnover pathway, with a concomitant release of NH₃ and CO₂. In this form, reduced nitrogen and carbon originating from the CNglcs could be utilized in primary metabolism. This alternative pathway might operate concurrently with the “conventional” bioactivation pathway, in which amygdalin and prunasin are hydrolyzed and decomposed into benzaldehyde and HCN; the latter is further detoxified through β-cyanoalanine into asparagine, aspartate and NH₃.

Based on the general alternative turnover pathway proposed by Pičmanová et al. (2015), we suggest three possible routes for the turnover of CNglcs in Prunus species (Figure 7), starting with the hydrolysis of amygdalin to prunasin. Then, in the first route, prunasin is further hydrolysed to prunasin amide and/or acid and NH₃. Prunasin acid is converted into prunasin anitrile or to β-D-glucose-1-benzoate, with a release of CO₂. In the second route, prunasin is converted directly into the corresponding anitrile with the release of NH₃ and CO₂. NH₃ as CO₂ produced in these proposed pathways may be channeled into primary metabolism. In a third route, prunasin is glycosylated to a prunasin pentoside that would also produce NH₃ and CO₂, when converted to prunasin anitrile pentoside. The latter could also be deglycosylated into prunasin anitrile.

Cyanogenic glucosides are biosynthesized from amino acids, therefore the plant must mobilize and transport these precursor substances to the sites where CNglcs are needed. Supply of nitrogen for the biosynthesis of CNglcs is especially important in young tissues, which are weaker than mature tissues and are in greater need of defense against pathogens and herbivores. On the other hand, at times where defense responses are less urgent, the plant can reuse nitrogen from CNglcs and redirect it into primary metabolism (Vries et al., 2017).

In Eucalyptus, it has been demonstrated that up to 20% of leaf nitrogen is stored in CNglcs, with the highest levels in young and reproductive tissues (Gleadow and Woodrow, 2000). In spring, coinciding with the flowering period, there was an important allocation of nitrogen to the reproductive tissues in detriment to the leaves to form CNglcs. The levels of these compounds decreased gradually during fruit development (buds – flowers – fruits).

Cyanogenic diglycosides may have additional functions as transport forms, pollinator attractants and germination inducers. In E. camphora trees, the highest levels of diglucosides were found in flower buds and expanded leaves (Neilson et al., 2011). Theoretically, the diglucosides are synthesized in the expanded leaves and then transported to the developing flower buds. The levels of cyanogenic diglucosides were much lower in immature fruits suggesting that nitrogen was remobilized and used during the flower development (Neilson et al., 2011).

As previously mentioned, HCN may be produced and metabolized during flower bud development, indicated by a decrease in CNglc levels. Past as well as recent studies have also shown that HCN may activate the flower bud and the flower opening in Lemna paucicostata and grapevine (Tanaka et al., 1983; Tohbe et al., 1998).

Interestingly, HCN has also been reported in releasing seed dormancy in orthodox seeds (Roberts, 1973; Roberts and Smith, 1977) by inducing the formation of Reactive Oxygen Species (ROS); ROS in turn activates a cascade involving Ethylene Response Factor 1 (ERF1), which leads to the production of germination-associated proteins (Oracz et al., 2009). Extensive literature describes the importance of the HCN in seed germination. Considering the common mechanisms regulating seed and bud dormancy, this process could be similar in endodormancy release (Taylorson and Hendricks, 1973; Bogatek et al., 1991; Flematti et al., 2013).

To obtain more information on the possible regulation of these processes, qRT-PCR analysis was performed on the sweet cherry samples. The expression levels of a selected number of genes were analyzed (Figure 8). In CNglcs biosynthesis: CYP79 and CYP71 (Figure 8A). In bioactivation: amygdalin and prunasin hydrolase (Figure 8B). In oxidation reactions: catalase and peroxidase (Figure 8C). In ethylene biosynthesis: SAM synthase, ACC synthase, and ACC oxidase (Figure 8D). In the detoxification pathway: L-3-cyanoalanine synthase (Figure 8E).

Both CYP79 genes displayed their highest level of expression after dormancy release, indicating that CNglcs biosynthesis takes place during early flower development in sweet cherry (Figure 8A). In buds of Japanese apricot (P. mume), CYP79A68 was the only examined cytochrome P450 monooxygenase encoding gene showing a substantial level of expression (Yamaguchi et al., 2014). Further Yamaguchi et al. (2014) reported that CYP79D16, but not CYP79A68, catalyzed the conversion of L-phenylalanine into E-phenylacetaldoxime. The second step in CNglcs biosynthesis is mediated by CYP71s (Sánchez-Pérez et al., 2008), such as CYP71AN24 and CYP71AP13 (Yamaguchi et al., 2014). In general, the expression of the two CYP71 encoding genes was transiently down-regulated shortly before dormancy release and subsequently increased again (Figure 8A). This is in accordance with the results for the CYP79s. Further CYP71AN24, but not CYP71AP13, catalyzed the conversion of E-phenylacetaldoxime into mandelonitrile (Yamaguchi et al., 2014). Hence, future studies have to reveal the functional properties and substrate specificities of CYP79s and CYP71s in sweet cherry to resolve the biosynthesis of prunasin in sweet cherry.

As previously mentioned, the degradation of CNglcs is initiated by β-glycosidases, in Prunus species called amygdalin hydrolase (AH) and prunasin hdyrolase (PH). Ah1 and Ph5 (Zhou et al., 2002) were examined in this study (Table 4) as they were the most similar characterized hydrolases between P. serotina and P. dulcis (Sánchez-Pérez et al., 2012). As shown in Figure 8B, both genes display transcriptional activity solely after dormancy release. In the case of Ph5, this fits well with the decrease of prunasin levels at around the same time point, indicating its degradation.

L-3-Cyanoalanine synthase (CAS) activity serves as an indicator for HCN release because of its essential involvement in HCN detoxification (Floss et al., 1965). After a transient peak, CAS transcription decreased and rised again during dormancy release and during flower development (Figure 8E).

Pathways involved in oxidative stress regulation have previously been shown to be active during dormancy release in several different perennials (Horvath, 2009; Cooke et al., 2012). In our study, catalase expression decreased slightly and then increased again just before dormancy was released (Figure 8C). Several studies found catalase activity to be affected by both natural and artificially induced bud break (Nir et al., 1986; Pérez and Lira, 2005; Amberger, 2013). Catalases are known to catalyze the conversion of H₂O₂ to water and oxygen (Chelikani et al., 2004). Thus, the inhibition of catalase gene transcription and enzyme activity by, e.g., HCN released from the cyanogenic glucoside hydrolysis could result in increased hydrogen peroxide levels. In this study, the subsequent up-regulation of the catalase gene after dormancy release might decrease H₂O₂ levels again, which is consistent with a steady decrease in H₂O₂ content after dormancy release found in flower buds of P. pyrifolia (Japanese pear) (Kuroda et al., 2002).

In addition to catalase, a range of peroxidases are able to reduce H₂O₂ to water and have been shown to be induced in response to oxidative stress during dormancy release in grape buds (Veitch, 2004; Keilin et al., 2007). The peroxidase gene examined in our study (Figure 8C) was most highly expressed at bud dormancy release, indicating that peroxidase functions mainly during the transition from dormancy to flowering in sweet cherry, which is similar to results acquired in Japanese pear (Bai et al., 2013). Differently regulated peroxidases during transition of dormancy release were observed in prior studies. For instance, in buds of Chinese cherry (P. pseudocerasus Lindl.), different peroxidase encoding genes were either down-regulated before, during and after dormancy release under natural conditions (Zhu et al., 2015). This suggests a pattern of alternating activities among a set of peroxidases that regulate oxidative stress during bud dormancy release. Peroxidases were found to be up-regulated in buds of peach and leafy spurge (Euphorbia esula L.) (Jia et al., 2006; Leida et al., 2010) and down-regulated in grapevine in regard to dormancy release (Pacey-Miller et al., 2003). The examined peroxidase gene in our study was down-regulated before dormancy release, which coincides with our observation of a decreased catalase expression, potentially giving rise to a transient increase in ROS. Subsequently enhanced expression of peroxidase and catalase encoding genes during and after bud dormancy release might then cooperatively reduce oxidative stress.

Transcript analysis of three key genes encoding enzyme involved in ethylene biosynthesis, namely S-adenosyl-methionine (SAM) synthetase, 1-aminocyclopropane-1-carboxylic acid (ACC) synthase and ACC oxidase were conducted and demonstrated that ACC synthase and ACC oxidase were initially expressed shortly before dormancy release (Figure 8D). Those results suggest that ethylene biosynthesis was initiated before dormancy release in sweet cherry. In grapevine, the effect of different temperatures and sampling dates on bud break and ACC content was studied, seeing that under low temperatures, bud break was associated with the promotion of ethylene biosynthesis (El-Shereif et al., 2005). Heat shock experiments demonstrated that ACC and ethylene accumulated toward dormancy release in grapevine (Tohbe et al., 1998). Transcription of the gene encoding ACC synthase was induced in flower buds in Japanese pear (Bai et al., 2013). Exogenous application of ACC has been reported to enhance dormancy release. The same effect was not observed upon exposure to ethylene (Iwasaki, 1980). Since hydrogen cyanide is formed in stoichiometric amounts with ethylene in the ACC oxidase catalyzed conversion of ACC, hydrogen cyanide is thought to be responsible for bud break in grapevine.