Miyabeacin: A new cyclodimer presents a potential role for willow in cancer therapy

Willow (Salix spp.) is well known as a source of medicinal compounds, the most famous being salicin, the progenitor of aspirin. Here we describe the isolation, structure determination, and anti-cancer activity of a cyclodimeric salicinoid (miyabeacin) from S. miyabeana and S. dasyclados. We also show that the capability to produce such dimers is a heritable trait and how variation in structures of natural miyabeacin analogues is derived via cross-over Diels-Alder reactions from pools of ortho-quinol precursors. These transient ortho-quinols have a role in the, as yet uncharacterised, biosynthetic pathways around salicortin, the major salicinoid of many willow genotypes.

For full structure determination by NMR, 3 was isolated via a larger scale extraction of S. miyabeana tissue and collection of the appropriate peak from semi-preparative HPLC. The 1 H NMR spectrum of 3 showed peaks relating to 34 coupled hydrogens (Table 1 and Supplementary Fig. 4). Four signals were observed between δ7.34-7.10 and matched those observed for the salicyl ring hydrogens in salicortin ( Supplementary Fig. 4). However, in 3 the integration of the aromatic peaks corresponded to 8 hydrogens suggestive of two such salicyl rings. This was confirmed by the presence of two pairs of J = 12 Hz doublet signals relating to the distinctive salicyl hydroxymethylene group (pair 1: δ5.40 and δ5.19; pair 2: δ5.38 and δ5. 16). Similarly, the molecule appeared to contain two separate β-glucoside moieties since characteristic doublet signals relating to the H-1′ anomeric hydrogen atoms were duplicated (δ5.09 and δ5.07) as were those corresponding to the glucosyl 6′-methylenes. Unlike the 1 H-NMR spectrum of salicortin 2 which showed aliphatic signals corresponding to the four methylene hydrogens of the HCC ring between δ2.94 and 2.51, the spectrum of 3 showed no signals in this region indicating a significant change in this part of the molecule. Similarly, signals from the olefinic hydrogens in the HCC-ring that appear in the salicortin 1 H-NMR spectrum at δ6. 27 and δ5.76 each as a doublet of triplets, were absent. In 3 these signals were replaced by four separate olefin signals between δ6.60 and 5.85 each integrating for one hydrogen, two appearing as double doublets and the others as simple triplets. A dienone arrangement was ruled out after examination of the 1 H-1 H COSY spectrum ( Supplementary Fig. 5) which clearly demonstrated that although two double bonds were present in the molecule, they were isolated from each other. Integration of the carbohydrate region (δ3.96-3.40) in the 1D spectrum suggested a total of sixteen coupled hydrogens. Of these, twelve could be accounted for in two glucose units leaving four unaccounted for. 13 C NMR data (Table 2 and Supplementary Fig. 6) confirmed the presence of forty carbon atoms in the molecule including two ketone carbonyls at δ198.6 and δ210.0 and two ester carbonyls at δ173.6 and δ173.2 whilst 13 C DEPT identified four non-aromatic methine signals, in addition to those of glucose (x2) and the two olefins indicated in the 1 H spectrum ( Supplementary Fig. 7). Given the molecular formula from accurate mass, the similarity in fragmentation pattern of the smaller m/z 421 fragment to that of salicortin, and the duplication of benzyl and glycosyl related NMR signals we postulated 3 was an unsymmetrical dimeric structure formed via condensation of two molecules of a salicortin analogue with dehydro-HCC rings. To satisfy the observed spectral data we concluded that the dimeric molecule had arisen via a [4 + 2] Diels-Alder cyclodimerisation reaction, between two "salicortenone", 4 ( Fig. 1) monomer units, one acting as the diene, and the second as the dienophile. The deduced structure of 3 has a diketo-1,4-ethenodecalin core, bearing pendant carboxy groups esterified with two salicin units and is shown in Fig. 1 and was named miyabeacin 3.
Re-analysis of the smaller fragment ions in the mass spectrum of miyabeacin ( Supplementary Fig. 1), showed that m/z 331.1034 corresponded to the formate adduct of salicin, 1, m/z 421.1140 to the retro-Diels-Alder product "salicortenone" 4, m/z 557.1300 to a mono-desalicylated dimeric moiety and the fragment ion m/z 217.0507, corresponding to the cyclo-dimer core structure (C 12 H 9 O 4 ) resulting from the loss of both carboxyl side chains. Final confirmation of the structure came from analysis of HSQC and HMBC NMR data ( Supplementary Figs. 8 and 9).  Fig. 1  The structure represents a completely new and unusual salicinoid. However, the aglycone has structural similarity to grandifloracin 5 (Fig. 1), a natural product first isolated from Uvaria grandiflora [(−) enantiomer] 20 and later from Uvaria dac [(+) enantiomer] by Awale et al., who demonstrated that the molecule had potent activity against human pancreatic cancer cells 21 . The structure of (+)-grandifloracin has been confirmed by both synthesis 22,23 and by X-ray crystallography 21 . Inspection of the published NMR data 21,23 relating to the olefin and bridgehead methine hydrogens from grandifloracin agreed well with our observed 1 H-NMR data of 3 ( Supplementary  Fig. 10). An important structural difference between miyabeacin 3 and grandifloracin 5 is the oxidation level at C-8 and C-21 (COOH vs CH 2 OH) and hence the orientation of the ester linkage of the attached aromatic ring. The [4 + 2] cyclodimerisation of substituted cyclohexadienones is a well-known reaction in synthetic chemistry and, aside from grandifloracin, there are a number of other examples of this reaction occurring in the natural world 24 . A key feature of the reaction, whether proceeding in vivo or in vitro is the exquisite regio-and stereo-selectivity leading to entirely endo-stereochemistry in the product.
We tested miyabeacin against a range of cancer cell lines. Initial cell viability assays were carried out using the MYCN-amplified neuroblastoma cell line UKF-NB-3, established from a stage 4 neuroblastoma patient 25 and the vincristine-resistant UKF-NB-3 sub-line UKF-NB-3 r VCR 10 (adapted to growth in the presence of vincristine 10 ng/mL). At a concentration of 20 µg/mL of miyabeacin, the cell viability, relative to non-treated cells, after 120 hours was 0% for UKF-NB-3 and 4.22 ± 2.89% for the vincristine resistant UKF-NB-3 r VCR 10 line. Initial IC 50 values were obtained from an expanded range of cell lines including those in neuroblastoma (UKF-NB-3), breast (BT-474 and MCF-7), oesophageal (COLO-680N) and ovarian cancers (COLO-704 and EFO-21) ( Table 3) and values ranged from 2.19-27.04 µg/mL. Three cell lines were selected and fully replicated IC 50 data was obtained (   (Table 1 and Supplementary Fig. 11). The presence of signals relating to benzyl and glucosyl moieties compared well with those observed in the 1 H-NMR spectrum of 3. Absence of the four olefin signals (δ5.91 to 6.59) previously observed in 3 was accompanied by a movement upfield of the four bridgehead hydrogens (δ3.43-3.63) to give a set of four signals at δ2.76, δ2.88, δ2.99 and δ3.12 each integrating for 2 hydrogens. The 1 H-NMR data suggested a further [2 + 2] intramolecular cyclization of the olefin units in 3 to give a "caged" structure which we have named miyabeacin B, 6. The cycloaddition of the double bonds in 3 to yield the cyclobutyl 'cage' in 6 now confers a 2-fold axis of symmetry resulting in a significant simplification of the 1 H-NMR spectrum for 6 relative to that of miyabeacin 3. This symmetry was also observed in the 13 C data ( Table 2) (Table 1 and Supplementary Fig. 16) suggested an analogous structure to the cyclodimer miyabeacin 3, although certain regions of the spectrum, including those relating to the benzyl and glucosyl groups, were no longer duplicated suggesting that one of each of these units had been lost. 1 H signals at δ6.63 and δ6.02 corresponded to those observed in 3 and related to the enone hydrogens, H-12 and H-13. Signals corresponding to the isolated olefin hydrogens at δ6.27 and δ5.94 were also present. These data and additional 1 H-1 H COSY correlations of these signals to 4 additional methine hydrogen atoms confirmed that the molecule retained the Diels-Alder "core" present in 3 (Supplementary Table 3 and S17). The NMR data were consistent with the MS data and suggested that compared to 3, compound 7 (MW 532, C 26 H 28 O 12 ) was missing a salicin side chain and a carboxy group. The remaining peaks in the MS spectrum (m/z 421.1120 and 467.1159), (Supplementary Fig. 15) were annotated as "salicortenone" 4 and its formate adduct and were believed to have arisen from a retro-Diels Alder fragmentation in the MS source. 13 C NMR showed 26 separate carbon signals including two ketone signals at δ213.3 and δ199.0 ( Table 2). The position of side-chain loss and decarboxylation was confirmed via extensive analysis of COSY, HSQC and HMBC correlation spectroscopy (Supplementary Table 3 and Figs. [17][18][19]. Key 1 H-13 C correlations were between H-10 and C-8, H-10 to C-14 and H-13 to C-9. This allowed placement of the carboxy-salicylglycoside moiety at C-9. However, no signals or correlations were evident to H-20, suggesting that hydrogen-deuterium exchange, through an ene-diol canonical form, had occurred with the NMR solvent. However, use of 100% D 2 O as solvent (Supplementary Table 3 Fig. 20). Additionally, an extra fragment ion at m/z 159.0458 corresponded to a fragment with a formula of C 10 H 7 O 2. 8 was isolated by fraction collection from HPLC and the 1 H NMR spectrum showed a downfield shift of the bridge olefin hydrogens (compared to 7) and also contained (2020) 10:6477 | https://doi.org/10.1038/s41598-020-63349-1 www.nature.com/scientificreports www.nature.com/scientificreports/ just two methine signals which were also well downfield of those seen in 7 (Supplementary Table 4 and Fig. 21). 13 C NMR data, derived from HSQC and HMBC experiments (Supplementary Figs. 22,23) indicated twelve aromatic carbons of which six were consistent with those observed for a salicyl glycoside (Supplementary Table 4).    www.nature.com/scientificreports www.nature.com/scientificreports/ Of the remaining aromatic C-signals, two were methines bearing hydrogens that were ortho-coupled (J = 8 Hz) to each other only, whilst the other four were quaternary, two of which were observed at δ144.5 and δ147.3 indicating hydroxyl substitution. The deduced structure, miyaquinol 8 ( Fig. 1) was confirmed by HMBC correlations to be an ortho-substituted di-phenolic ring fused to a bicyclo-octene moiety bearing a ketone at C-10 and the carboxysalicyl glucoside and hydroxyl groups at C-9.
[4 + 2] dimers are not confined to Salix miyabeana, and accumulate over the growth cycle.
Juvenile leaf and stem tissue from 26 field-grown Salix species were profiled for the dimeric salicinoids (Supplementary Table 5). Significant levels (>40 mg/g d.w.) of dimeric analogues were found in 3 accessions (NWC941, NWC885 and NWC837) of S. miyabeana Seemen, and also in 2 accessions (NWC577 and NWC592) of S. dasyclados Wimm. Lower levels were found in 2 additional S. dasyclados Wimm. accessions (Fig. 3). Whilst ratios between miyabeacin 3, miyabeacin B 6 and miyabeanol 7 varied, all compounds were found in leaf tissue. In stem tissue, only 3 and 6 were detected. Of lines that produced dimeric compounds, there was a strong correlation in the concentrations of miyabeacin 3 and miyabeacin B 6 (r 2 > 0.95, p < 0.001) in both leaf and stem samples suggesting a direct biosynthetic relationship. A slightly lower correlation (r 2 = 0.86, p < 0.001) between the concentration of miyabeacin 3 and miyabeanol 7 in leaf tissue ( Supplementary Fig. 24), was also observed. Comparison of uHPLC-MS peak areas of miyabeanol 7 and 8 indicated a high degree of correlation (r 2 = 0.982, p < 0.001) and a ratio of approximately 9:1 with 7 being the major of the two products ( Supplementary Fig. 25).
To investigate the onset of dimeric salicinoid production and determine the temporal relationship of the cyclodimerisation with the production of salicortin and salicin, S. dasyclados tissue was sampled from 3 to 60 days after budburst (phenological stage 10) 19 from an array of pot-grown clones generated from cuttings and grown in controlled environment under long days. Destructive sampling was carried out daily at the same time from 3-10 days (T3-T10) and subsequently at 12, 15, 30, 45 and 60 days after bud burst (Supplementary Fig. 26). Compounds 3, 6 and 7 were quantified from 1 H-NMR data alongside levels of salicin 1 and salicortin 2, saligenin (salicyl alcohol) and catechol by integration of compound specific NMR signals. In leaves, miyabeacin 3 and miyabeacin B 6 were present in very young T3 shoots at levels of 36 µmoles/g d.w and 1.9 µmoles/g d.w respectively (Fig. 4A). Levels of these compounds increased from T3 to T8 and then remained level until T15. Levels increased further at T30 and T45 until finally falling at T60. A correlation coefficient of r = 0.87 (p < 0.001) indicated that production of these two dimeric entities were highly correlated (Fig. 4B) and confirmed the hypothesis that 6 is derived from 3 from a [2 + 2] cycloaddition reaction. Levels of salicortin 2 were in the same range (20-50 µmoles/g d.w) but the trajectory was less correlated with 3 and 6 with a coefficient of r = 0.58. Levels of, miyabeanol, 7 were inversely correlated (r = −0.63) with 3, being highest at T3 (69.9 µmoles/g d.w.) and falling to 10.4 µmoles/g d.w. by T60. Salicin 1 levels were also highest at T3 (63.8 9 µmoles/g d.w.), decreased gradually across the time course and were highly correlated (r = 0.96) with miyabeanol, 7 concentrations. Similarly, catechol and saligenin levels also decreased from T30 to T60 and although observed at lower concentrations were highly correlated with salicin concentrations across the experiment (both having a correlation coefficient of r = 0.98 to salicin). The data indicated a relationship of miyabeanol, 7 to salicin 1, saligenin and catechol which was in contrast to the increasing levels of miyabeacin, 3 and salicortin 2.
In stem tissue from the same plant samples the concentration profiles (Fig. 4C) of all metabolites mirrored that observed in leaf tissue with the exception of miyabeanol, 7 which could only be detected in the very juvenile stem tissue of young T3 and T4 shoots. It is possible that 7, is only produced in leaves and that at the very early time points T3 and T4, the stem and leaf are not yet fully differentiated. the cyclodimerization process is heritable and substrate driven giving rise to "cross-over" Diels-Alder products. S. miyabeana and S. dasyclados have been used as breeding parents for the generation of willow lines with improved biomass traits. The commercial variety "Terra Nova", a hybrid arising from S. miyabeana, S viminalis and S. triandra (Supplementary Table 6) was analysed by LC-MS ( Supplementary Fig. 27) which confirmed the presence of cyclodimeric analogues 3, 6 and 7 in juvenile leaf tissue. In line with the parent, S. miyabeana, cyclodimers were also present in the stem tissue with the exception of 7. A more striking example of the heritability of the cyclodimer production was observed for the biomass variety "Endurance", a hybrid of S. rehderiana and S. dasyclados (Supplementary Table 6), the former a species of willow that produces acetylated salicinoids. The LC-MS profile ( Supplementary Fig. 28) again contained compounds 3, 6 and 7 but also contained related compounds all bearing at least one acetyl moiety.  Fig. 29A). Isolation via HPLC peak collection and collection of 1 H-NMR spectra indicated a 1:1 mixture of monoacetylated miyabeacin isomers (9a and 9b) www.nature.com/scientificreports www.nature.com/scientificreports/ δ2.137. Appearance of 2 double doublets (J = 8.0, 9.6) at δ5.00 and δ4.97 placed the acetyl moiety in the glucose groups at 2′ in one isomer and at 2′′ in the alternate compound (Supplementary Table 7). Substitution on the glucosyl ring now produced separated signals across the 1 H spectrum for H-1′/1′′, H-3′/3′′, H-4′/4′′ and H-5′/5′′. Additionally, benzyl hydrogens (H 2 −7) showed an upfield shift due to acetylation at the 2′-position and now appeared at δ5.09/ 5.03 in 9a while corresponding H 2 −22 moved to δ5.10/ 5.03 in 9b when acetylation occurred at the 2′′-position. A full 1 H and 13 C spectral assignment (Supplementary Table 7  , the former corresponding to a benzoate moiety and the latter to a salicyl moiety. The data were suggestive of 16a/16b, novel mono-benzoylated derivative of miyabeacin 3. The peak was isolated by repeated injection into an HPLC system and the structure characterised by 1 H-NMR (Supplementary Fig. 33). This data, via comparison to the NMR spectrum of 3, confirmed the presence of a dimeric compound. Additional peaks at δ8.06, δ7.70 and δ7.54 were characteristic of a benzoate group but a doubling of most peaks indicated a 1:1 mixture of isomers which could not be separated further. A set of multiplet peaks appearing at δ5.25 confirmed the benzoyl substitution at the 2′-position of the glucoside in the first isomer (16a) and the 2′′-position in the second isomer (16b). This was further confirmed in the 1 H-1 H -COSY spectrum (Supplementary Fig. 34). www.nature.com/scientificreports www.nature.com/scientificreports/ A further example, extending the range of substituted dimeric compounds was seen in the LC-MS analysis of a willow breeding line (RR10147) developed as part of a biomass improvement programme at Rothamsted Research. This hybrid line included S. dasyclados (NWC577) in both parents [RR07187 (NWC944 S. glaucophyloides × NWC577 S. dasyclados "77056") × RR07188 (NWC944 S. glaucophyloides × NWC577 S. dasyclados "77056")] as well as S. glaucophyloides (NWC944). In the Total Ion Chromatogram of the negative ion mode LC-MS data ( Supplementary Fig. 35) salicortin 2, 2′-O-acetylsalicortin 14 and tremulacin 15 appeared as major peaks. Given that this cross has generated a hybrid capable of producing both acetylated and benzoylated salicinoids alongside salicortin it followed that associated dimeric analogues would also be expected to be formed via a matrix of cross-over reactions involving the three corresponding dienones 2,12 and 17. This was indeed the case with miyabeacin 3 appearing at 25.03 min, 2′/2′′-O-acetylmiyabeacin 9a/9b appearing at 26.90 min and 2′/2′′-O-benzoylmiyabeacin 16a/16b appearing at 30.95 min. A further intriguing peak was observed at 32.48 min which showed an ion at m/z 989.2617, corresponding to a formula of C 49 H 49 O 22. Although there was insufficient for isolation, the MS was suggestive of the predicted miyabeacin analogue bearing both an acetyl and benzoyl substitution.

Discussion
The identification of miyabeacin 3 and its analogues in Salix spp. adds further examples to the group of natural products that are formed from intermolecular Diels-Alder, cyclodimerisation reactions of ortho-quinonoids 24 . These precedents include dimers of terpenoid and polyketide intermediates, but also the phenolic-derived analogue, grandifloracin 5 which bears strong structural similarities to the aglycone of miyabeacin, although it is important to note that, relative to 5, the side-chain ester functions of 3 are inversely orientated, and also that in the case of miyabeacin the reaction involves highly polar glycosylated reactants. The obligate substrate for the formation of miyabeacin by cyclodimerisation in willow is the substituted ortho-quinol, salicortenone 4 which has a dienone unit that behaves as both a diene and a dienophile, the latter reacting through the distal double bond of the dienone (C-10 and 11) as shown in Fig. 5. This reaction, for these types of orthoquinol reactants, is known to be exquisitely regio-and stereo-specific to give only endo-products with no known natural compounds resulting from either exo-processes or reaction at the alternate double bond of the dienophile (in this case C12 and 13, Fig. 5.) 24,26 . In this work we have also identified a further dimeric natural product -the 'cage' compound, miyabeacin B 6 that arises from an additional [2 + 2] photo-annelation reaction of the [4 + 2] Diels Alder product. There are also precedents, in synthetic examples, for this conversion 24 which again confirms the endo selectivity of the preceding [4 + 2] reaction as the double bonds in the corresponding exo products are not aligned for [2 + 2] cage formation. The natural [4 + 2] reaction occurs spontaneously at ambient temperature and, as yet, there has been no evidence provided, for any enyzme directly catalysing intermolecular Diels-Alder reactions 24 . However, we have now presented strong evidence from analysis of breeding parents and progeny that the ability to produce the cyclodimers is a heritable trait that is genetically encoded in S. miyabeana and S. dasyclados and their progeny. Furthermore, the observation of cross-over [4 + 2] cycloaddition reactions when Salix genotypes for example, S. rehderiana that produces 2′-acetylsalicortin 14 and S. rossica that produces 2′-benzoylsalicortin (tremulacin) 15 are crossed, respectively, with S. dasyclados and S. miyabeana, provides further compelling evidence for the inheritance of the Diels-Alder capability. An exquisite example of the combinatorial nature of the reactions was observed in the complex hybrids that contain both 2′-acetyl and benzoyl substrates as well as the [4 + 2] capability from S. dasyclados, where all possibilities of cross-over dimerisations were observed. These results also indicate that the glucose and substituted glucose moieties in miyabeacin and its analogues must be already present in the dienone substrates, indicating that glycosylation, and 2′-acylation of that glucoside, are earlier steps in the salicinoid pathway.
Apart from the unlikely possibility of the direct involvement of a gene-encoded protein that binds 2 molecules of the monomer salicortenone 4 (and/or its acylated analogues) and facilitates the cyclodimerisation, the most likely explanation of the heritability of the production of miyabeacin and its relatives is that in S. miyabeana and S. dasyclados there is genetic control of the production or accumulation of the dienones such as 4. This novel trait/gene is not present or not functional in willow lines that only express the common pathway to salicortin and analogues such as tremulacin. The production of the cross-over [4 + 2] products bearing acyl groups when two analogous salicortenones are co-biosynthesised supports the argument that the pool size of salicortenones is genetically controlled. Given that the spontaneous [4 + 2] dimerisation is under kinetic control, concentrations of precursor in planta are a major determinant. Many willow species make salicortin and given the structural similarity of salicortenone to salicortin it is likely that they are closely related biosynthetically. Two biosynthetic scenarios are possible as shown in Fig. 6. The first and more straightforward possibility is that salicortenone 4, and hence miyabeacin 3, are metabolites of salicortin 2, formed from an oxidative dehydrogenation process (e.g. via a P450 reaction) inheritable from S. miyabeana or S. dasyclados. In this scenario an extra gene, present in S. miyabeana or dasyclados, is carried into hybrid progeny where, depending on the species, a variety of salicortin derivatives are presented to the new enzyme for conversion to the dienone and hence the dimers. Endurance inherits the [4 + 2] process from S. dasyclados and the 2′-acetylation trait from S. rehderiana. The observation that mono-acetylmiyabeacins 9a and 9b are produced in a 1:1 ratio indicates that pools of "salicortenone" 4 and "acetylsalicortenone" 12 are both available for "cross-over" dimerization reactions, and that, in planta, the acetylation of the glucose moieties occurs before, rather than after, the Diels-Alder reaction. The dienones 4 and 12 could not be identified in the samples, and thus remain obligate but undetected biosynthetic intermediates. The biosynthetic precursor or catabolite relationship of salicortin 2 to the structurally close dienone 4 is also unclear, but relative levels of salicortin 2 and 2′-acetylsalicortin 14 in the hybrid may be expected to predict the ratios 3, 9(9a + 9b) and 10 produced in the cross-over reaction. Salicortin 2 and 2′-acetylsalicortin 14 were quantified directly from 1 H-NMR data of the plant extracts and occurred in a ratio of 6:1 in stems and 1:2 in leaves. Visual inspection of LC-MS analysis of Endurance ( Supplementary Fig. 28) shows that the population of cyclo-dimers containing 0,1 or 2 acetyl groups on the glucose moieties do reflect the relative levels of (2020) 10:6477 | https://doi.org/10.1038/s41598-020-63349-1 www.nature.com/scientificreports www.nature.com/scientificreports/ salicortin 2 and 2′-acetylsalicortin 14 in the two tissues. However, the ratios of 3, 9 and 10 are not totally consistent with the notion of simultaneous formation and free mixing of pools of salicortenone 4 and 2′-acetylsalicortenone 12, which would predict that more 10 than 3 would be produced in leaves where acetylsalicortin 14 predominates over salicortin 2. The observed ratios of cross-over dimers suggest that the affinity of the oxidative enzyme proposed in scenario 1 (Fig. 6.) may be slightly different for salicortin 2 and the acetyl analogue 14 (and thus also 15), which can explain why the observed ratio of 3, 9a + 9b and 10 differs from that expected by the relative concentrations of 2 and 14, particularly in leaf tissue.
Here we suggest that a "normal" intimate coupling (channelling) between the enzyme(s) that produces 4 and a reductive enzyme, that potentially converts 4 to salicortin via a 1,4-reduction process, is impaired in S. miyabeana and S. dasyclados, resulting in pooling and hence spontaneous dimerisation of a significant proportion of 4. However, both S. miyabeana and S. dasyclados still produce salicortin 2, and thus the reduction process cannot be completely impaired. In this biosynthetic scenario it is this partial loss of reductive function is passed on to the hybrids, a situation that is genetically more complex, but against a background of polypoid genomes, may have more to do with titre of fully coupled and functional reductases. The time-course data (Fig. 4) suggests that salicortin and miyabeacin production in planta are linked, but this would be so in both scenarios 1 and 2 (Fig. 6).
The origins of miyabeanol 7 and the minor related miyaquinol 8 which occur only in leaves, and very young green shoots at T3 and T4 (Fig. 4) is also not completely clear. Chemically, 7 and 8 could arise, in the plant or on sample processing, from hydrolytic removal of a salicin 1 moiety from miyabeacin 3 with concomitant decarboxylation (for 7) or aromatisation via loss of HCOOH (for 8). We cannot rule out non-enzymic hydrolysis completely but the lack of 7 in mature stem tissue that accumulates 3 indicates that the hydrolysis is not artifactual due to processing, but would need to be specific to photosynthetic tissue, suggesting involvement of a leaf carboxyesterase. In vitro, miyabeacin 3 was found to be stable, in neutral aqueous solution for several months, but underwent hydrolysis in mild alkaline solution to give salicin, miyabeanol 7, miyaquinol 8 and catechol. Thus, another explanation for spatial differences in the occurrence of 7 in planta, could be a pH difference in the two tissues. The strong correlation of miyabeanol, salicin and catechol accumulation in the leaf time-course (Fig. 4) indicates that a significant portion of salicin production in S. miyabeana and S. dasyclados in this tissue occurs through the turnover of miyabeacin via miyabeanol. However, in mature stem tissue, where miyabeanol is absent, this route is perhaps less important, and the salicin may arise from the 'normal' route shared with other Salix genotypes. This 'normal' route to salicin is not clear but may involve direct synthesis from salicylaldehyde 28 or via cleavage of esters such as salicin-7-benzoate and/ or downstream products such as salicortin 17 . The degree of relocation of molecules from leaf to stem is an unknown factor and it should also be noted that salicortin is also susceptible to hydrolysis in vitro, the major products being salicin and catechol 29,30 . Thus, in summary, a network of potential routes to salicin exists, which in S. miyabeana and S. dasyclados appears to be supplemented by a considerable contribution from miyabeacin catabolism.
The pharmaceutical activity of salicin 1 is well known, but the potential pharmacology of miyabeacin 3 is perhaps much wider. Structurally, it contains two salicin groups that give it potential 'double dose' anti-inflammatory and anti-thrombolytic activities associated with salicin and aspirin. However, our results reporting the activity of miyabeacin 3 against a number of cancer cell lines including cell lines with acquired drug resistance, adds further evidence for the multi-faceted pharmacology of willow. Of particular note is the activity against neuroblastoma cell lines. Overall survival rates are below 50% and it represents the most frequent extracranial solid childhood tumour 31 . With resistance acquisition being a significant issue in neuroblastoma 32 , new drugs with novel modes of action are required and miyabeacin perhaps offers a new opportunity in this respect. www.nature.com/scientificreports www.nature.com/scientificreports/ were separated prior to milling to a fine powder (Retsch Ultra Centrifugal Mill ZM200, Retsch, UK). Milled tissue was maintained at −80 °C until analysis. A voucher specimen has been retained and is available on request. Additional sampling for comparative miyabeacin assessment in other species took place from the same collection within 2 weeks. Tissues were harvested at the same time of day, and were harvested and processed using identical protocols Generation of plant tissue for detailed timecourse. 15 cm cutting stems of S. dasyclados (NWC577), harvested from the National Willow Collection, Rothamsted Research, UK during dormancy (December 2014), were removed from cold storage (−4 °C), fully defrosted for 6 hours, then soaked overnight in tap water. 50 Cutting stems were planted at half-depth in individual pots (15 cm Ø, 15 cm tall, volume 1.55 L) consisting of a 50:50 v/v of perlite and Rothamsted Standard Compost Mix (75% Medium grade (L&P) peat, 12% Screened sterilised loam, 3% Medium grade vermiculite, 10% Grit (5 mm screened, lime free) plus 3.5 kg Osmocote Exact 3/4 month per m 3 ). Plants were grown under controlled environment conditions, set to achieve an average of 600 μmol·m −2 ·s −1 measured at plant growth height (2.5 m from light source). Plants were grown under 14/10 hr day/nights at 18 °C/10 °C and 60%/90% relative humidity. Plants were observed daily for signs of budburst and tagged (T0) when they achieved this (normally 7 to 10 days from planting). "Budburst" corresponded with the Weih definition of bud burst (stage 3) 33 and stage 10 as suggested by Saka and Kuzovkina 19 10 whereby "green tip elongates but leaves till remain in a tight cluster (>5 mm)". Tissue from multiple shoots was cut with scissors and immediately snap-frozen in liquid nitrogen at the following number of days after budburst (T): 3,4,5,6,7,8,9,10,12,15,30,45 and 60. For T3-T5 stages tissue from 2 plants (multiple stems) were pooled for each biological replicate. For the remaining timepoints, a single plant's shoots were harvested. Three biological replicates per timepoint were harvested. Tissue was kept at −80 °C prior to lyophilisation. Stem and leaf tissues were separated prior to milling to a fine powder using a pestle and mortar. Milled tissue was maintained at −80 °C until analysis.