Formation of lignans (-)-secoisolariciresinol and (-)-matairesinol with Forsythia intermedia cell-free extracts.

In vivo labeling experiments of Forsythia intermedia plant tissue with [8-14C]- and [9,9-2H2,OC2H3]coniferyl alcohols revealed that the lignans, (-)-secoisolariciresinol and (-)-matairesinol, were derived from two coniferyl alcohol molecules; no evidence for the formation of the corresponding (+)-enantiomers was found. Administration of (+-)-[Ar-3H]secoisolariciresinols to excised shoots of F. intermedia resulted in a significant conversion into (-)-matairesinol; again, the (+)-antipode was not detected. Experiments using cell-free extracts of F. intermedia confirmed and extended these findings. In the presence of NAD(P)H and H2O2, the cell-free extracts catalyzed the formation of (-)-secoisolariciresinol, with either [8-14C]- or [9,9-2H2,OC2H3]coniferyl alcohols as substrates. The (+)-enantiomer was not formed. Finally, when either (-)-[Ar-3H] or (+-)-[Ar-2H]secoisolariciresinols were used as substrates, in the presence of NAD(P), only (-)- and not (+)-matairesinol formation occurred. The other antipode, (+)-secoisolariciresinol, did not serve as a substrate for the formation of either (+)- or (-)-matairesinol. Thus, in F. intermedia, the formation of the lignan, (-)-secoisolariciresinol, occurs under strict stereochemical control, in a reaction or reactions requiring NAD(P)H and H2O2 as cofactors. This stereoselectivity is retained in the subsequent conversion into (-)-matairesinol, since (+)-secoisolariciresinol is not a substrate. These are the first two enzymes to be discovered in lignan formation.

Like the closely related polymeric lignins, lignans have been isolated from all parts of plant material (roots, leaves, stems, bark, etc.) but are mainly located in woody tissue, particularly heartwood (6-9). Currently, we have no knowledge regarding the actual site of lignan formation (biosynthesis) and the subcellular location where they are initially deposited or stored. It is often assumed that lignans are deposited first in the vacuole and are then ultimately secreted into the cell wall following vacuole collapse. This has never been rigorously proven.
In terms of their biosynthetic pathways and structures, lignans and lignins are products of the shikimate/chorismate and phenylpropanoid pathways, and both are structurally related. Many substructures in lignins contain the structural elements of isolated lignans.
Lignans and lignins, however, apparently differ in one fundamentally important aspect, namely optical activity. For the most part, dimeric lignans (e.g. secoisolariciresinol 1, pinoresinol 5, matairesinol 2, and podophyllotoxin 7) are optically active (1, 2), whereas isolated lignins are not. It is perhaps significant that higher oligomeric forms of lignans (trimers, tetramers, etc.) typically have only very small [ a ]~ values ( 5 ) . Indeed, the exact point of demarcation between oligomeric lignans and lignins is not well defined.
The optical rotation of a particular lignan can vary with plant source; e.g. Forsythia suspensa (10, 11) contains (+)pinoresinol 5a, whereas Xunthoxylum ailanthoides (12) has the (-)-enantiomer 5b. No satisfactory explanation has been proffered to account for this stereochemical control leading to optical activity, other than that the reaction is somehow enzymatically mediated. Such control is not possible via intercession of a typical peroxidase/H202-catalyzed reaction, a reaction often implicated in lignin synthesis (13).
Surprisingly, the biosynthesis of lignans has been a neglected area, even for medicinally important compounds such as podophyllotoxin 7, a chemical precursor for the drugs etoposide and teniposide in cancer chemotherapy (14,15). Indeed, not a single enzymatic step in the initial coupling of monomers, or any of the subsequent modifications (oxidations, ring closures, etc.), has ever been reported. This is all the more surprising because of the close chemical relationship between lignans and lignins.
In spite of substantive efforts (16, 17), unambiguous proof of the exact chemical nature of the phenylpropanoid monomers undergoing coupling to afford the lignan dimer skeleton had not been obtained. From our standpoint, two possibilities were under consideration: the lignans, matairesinol 2, arctigenin 3, arctiin 4, and podophyllotoxin 7 , could be formed alcohol^ and a hydroxycinn~ic acid (e.g. ferulic 10 or sinapic 11 acid). Alternatively, their formation could arise via direct coupling of either two monolignols or two hydroxycinnamic acids, with subsequent transformations occurring post-coupling. For example, secoisolariciresinol 1, pinoresinol 5, and epipinoresinol 6 could arise via direct coupling of the two monolignol molecules, coniferyl alcohol 8.
It must be emphasized that this uncertainty, as regards identity of the phenylpropanoid monomer(s) undergoing coupling, was a key issue, since none of the possibilities described above could be ruled out. Herein, we describe the direct coupling of two coniferyl alcohol 8 moieties affording only ~-~-secoisolaric~resinol lb, which is then stereoselectively converted into (-)-matairesinolZb. These conversions have been demonstrated using cell-free preparations from F. intermedia.

RESULTS AND DISCUSSION
The first goal of our research was to identify the key enzymatic reaction affording entry into the specialized biosynthetic pathway to the F o r s~t h~ lignans. This required identification of (i) the phenylpropanoid monomer(s) undergoing coupling (Le, the substrate or substrates); (ii) the type of enzymatic coupling reaction (oxidative or reductive); and (iii) the immediate coupling product and its stereochem-~ ' The "Experimental Procedures" are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press. istry. For a molecule such as matairesinol 2, its formation could occur either by coupling of one molecule of coniferyl alcohol 8 and one molecule of ferulic acid 10 followed by spontaneous lactone formation or via direct coupling of two coniferyl alcohol moieties to afford secoiso~ariciresinol 1 with subsequent dehydrogenation to give matairesinol2. Alternatively, ferulic acid 10 or coniferaldehyde could serve as immediate precursors. Based on structural considerations, we rationalized that the initial coupling product was either secoisolariciresinol 1 or matairesinol 2, and both lignans were obtained in racemic (+)-form by total synthesis. (+I-Matairesinols 2a/2b were formed using the method of Brown and Daugan (21) with the following exception: reduction of methyl 2-carboxymethyl-3-(4-hydroxy-3-methoxypheny1)propionate was carried out in 38.7% yield using a reducing agent, made in situ from nbutyllithium and diisobutyllithium aluminium hydride, rather than Ca(BH4)2, which, in our hands, consistently gave low yielding reactions. (~)-Secoiso~ariciresinols l a / l b were obtained by LiA1H4 reduction of (+)-matairesinols 2a/2b. Each racemic Iignan was resolved into its separate enantiomeric forms following passage through a Chiralcel OD column (Figs. 2A and 3A). (-1-Secoisolariciresinol l b was synthesized from (-~-rnatairesinolZb as above (Fig. 2B).
With a method to rapidly determine chirality, we next examined F. intermedia plant extracts to establish the optical purity of the secoisolariciresinol 1 and rnatairesinol2 present. Each lignan was isolated from methanol extracts of F~ intermedia stems. Matairesinol2 was relatively plentiful (1.05 mg g" dry plant tissue), whereas secoisolariciresinol 1 was less abundant ((0.036 mg g-' dry plant tissue). Chiral HPLC2 The abbreviations used are: HPLC, high performance liquid chromatography; m.pt, melting point; lit.m.pt., literature melting point; THF, tetrahydrofuran; TLC, thin layer chromato~aphy. analysis of both lignans (before recrystallization) revealed only the presence of the (-)-, and not (+)-, antipodes (Figs. 2C and 3B). This suggested that only the (-)-form was being synthesized in vivo, although a rapid interconversion of (+)into the (-)-forms, or into other metabolites, could not be ruled out.
Attention was next directed to establishing the chemical identity of the phenylpropanoid moiety undergoing coupling. In the first instance, [8-14C]coniferyl alcohol (1.30 mg, 23 KBq mg") was administered to F. intermedia shoots. Following its metabolism for 3 h, tbe plant material was homogenized, with unlabeled (f)-secoisolariciresinols l a / l b (100 pg) added as radiochemical carriers. The lignans were isolated as described in the Miniprint. First, secoisolariciresinol 1 and matairesinol2 were separated by reversed phase HPLC, using both radiochemical and UV detection. In this way, it was established that [8-'4C]coniferyl alcohol had been incorporated into secoisolariciresinol 1 (0.3%) and matairesinol 2 (1.8%), respectively, based upon coincidence of radioactivity peaks with eluted lignans. Next, secoisolariciresinol 1 and matairesinol 2 were collected (by HPLC separation) and subjected to chiral HPLC analysis. As can be seen from the radiochemical elution profile (Fig. 2 0 ) , only radiolabeled (-)secoisolariciresinol l b was detected in vivo. (Note that the UV elution profile shows the presence of both (+)-and (-)forms since unlabeled (&)-secoisolariciresinols l a / l b were added as radiochemical carriers.) In a similar manner to secoisolariciresinol 1, [8-'*C]coniferyl alcohol was only incorporated into (-)-matairesinol 2b, as evidenced by the radiochemical elution profile (Fig. 3C). (Again, the UV profile of matairesinol 2 shows the presence of both (+)-and (-)antipodes due to the addition of unlabeled carrier for chiral HPLC analysis; the large preponderance of the (-)-form reflects the amount of naturally occurring (-)-matairesinol 2b already present in F. intermedia tissue.) These experiments did not, however, prove that coniferyl alcohol 8 had been incorporated intactly into either lignan; enzymatic conversion of this alcohol to the acid or aldehyde could have occurred prior to coupling. Clearly, this uncertainty could be resolved by administration of [9,9-2H2,0C2H3] coniferyl alcohol to F. intermedia plant tissue actively synthesizing the lignans, (-)-secoisolariciresinol l b and (-)-matairesinol2b. If intact incorporation of coniferyl alcohol 8 occurred, then the (-)-secoisolariciresinol l b and (-)-matai-resinol2b formed de novo would contain 10 and 8 deuterium atoms, respectively. This could be proven by mass spectrometry. If, however, oxidation to the aldehyde or acid occurred prior to coupling, then the C9 position of the monomer would contain either one or no deuterium atom.
[9,9-*H2,0C2H3]Coniferyl alcohol obtained by total synthesis, as described in the Miniprint, was administered to excised F. intermedia shoots (0.59 mg/shoot) which were then allowed to metabolize for 3 h. Following this period, the lignans, (-)secoisolariciresinol l b and (-)-matairesinol 2b were isolated from F. intermedia, but without addition of unlabeled carrier,  Enzymatic formation of (-)-matairesinol2b from (&I- Standard assay conditions are described under "Experimental Procedures" and differ only in choice of NADP or NAD as shown. Protein content was 2.0 mg/ml. *Control experiments refer to the complete assay with either omission of cofactors or with denatured enzyme (boiled for 5 min). One other control was carried out, using the complete assay (with NADP) but with a reaction period of 10 s. In this experiment, the incorporation of radioactivity into (-) -matairesinol 2 b was 0.03%. 'H2,0C2H3]coniferyl alcohol molecules without prior C, oxidation. This is because signals at mlz 372 (M' + 10) and 354 (M' + 10, less H20) prove that the newly formed (-)-secoisolariciresinol l b contains ten deuterium atoms. Additionally, the peak at m/z 140, corresponding to a fragment derived from benzylic cleavage, reveals that the methoxyl group was fully deuterated. Formation of (-)-secoisolariciresinol l b can, therefore, occur only via coupling of two intact coniferyl alcohol 8 moieties.
Comparison of the mass spectrum of synthetic matairesinol 2 to that obtained following [9,9-ZH2,0C2H3]coniferyl alcohol feeding to F. intermedia was also informative (see Table I). As shown in Fig. 4E and Table I, unlabeled (-)-matairesinol 2b has two main signals at m/z 358 (M') and at 137 (derived from cleavage of the benzylic fragment). On the other hand, the (-)-matairesinol 2b isolated from F. intermedia previously treated with [9,9-2H2,0C2HHs]coniferyl alcohol gave signals at m/z 366, 358, 140, and 137 (Table I) z 358 and 137 again correspond'to natural abundance (-)matairesinol 2b, whereas the small signals at m/z 366 and 140 suggest that eight deuterium atoms had been incorporated, six of which were associated with the two methoxyl groups. (The relatively low intensities of the deuterated peaks are a consequence of unlabeled (-) -matairesinol2b previously accumulated in F. intermedia tissue. This is in contrast to that observed for (-)-secoisolariciresinol lb.) Having established that both halves of the (-)-secoisolariciresinol lb and (-)-matairesinol 2b molecules were derived from coniferyl alcohol 8, it was next of interest to determine whether (-)-matairesinol 2b was formed in viuo by direct dehydrogenation of (-)-secoisolariciresinol lb. To answer this question, (&)-[Ar-"Hlsecoisolariciresinols (17 KBq mg-') were synthesized (from unlabeled synthetic material by exchange with CF&Oi'H) and administered to F. intermedia plant tissue. After a 3-h metabolism, matairesinol 2 was isolated and subjected to reversed phase HPLC. Analysis of the resulting radiochemical elution profile revealed that the incorporation of (+)-[Ar-"Hlsecoisolariciresinols into matai-resin01 2 was 0.94%. The isolated [Ar-"Hlmatairesinol was subsequently subjected to chiral HPLC analysis, which demonstrated that only the (-)-antipode 2b was radiolabeled (Fig. 30). No radioactivity was detected in (+)-matairesinol 2a. (Note that the UV profile shows the presence of both enantiomers due to the addition of unlabeled (+)-matairesinols 2a/2b for chiral HPLC analysis.) These sets of experiments, therefore, suggest the following sequence of events in Go: coupling of two coniferyl alcohol 8 molecules to afford (-)-secoisolariciresinol lb and subsequent dehydrogenation to give (-)-matairesinol2b.
Our next objective was to determine whether such transformations (i.e. coupling and dehydrogenation) could be demonstrated in uitro using cell-free extracts from F. intermedia. Thus, incubation of [8-'4C]coniferyl alcohol with F. intermedia cell-free extracts for 1 h at 30 "C was carried out next (24). Following a series of experiments with appropriate cofactors (i.e. H202 and NAD(P)H), it was found that secoisolariciresinol 1 formation only occurred in the presence of H20, (0.4 mM) and NAD(P)H (4 mM). Subsequent chiral HPLC analysis of the isolated lignan revealed only formation of (-)-secoisolariciresinol lb, and not its (+)-antipode la. The rate of formation of (-)-secoisolariciresinol lb was 15.9 nmol h-' mg-' protein. Significantly, no formation of (-)secoisolariciresinol lb was observed when either cofactor was omitted (NAD(P)H or H20,) or when the enzyme was denatured (boiled 5 min). To further confirm that the enzymatic product was indeed (-)-secoisolariciresinol lb, [9,9-"HZ,0C2H3]coniferyl alcohol (6.97 mg) was incubated with the cell-free extract, in the presence of NADPH and Hz02. The enzymatic product was confirmed to be (-)-[2H,0]secoisolariciresinol by comparison of its mass spectrum with that of natural abundance (+)-secoisolariciresinols la/lb (24). It can, thus, be concluded that in this species, coupling of (-)-secolsolariciresinol m (-)-matalreslnol 2p coniferyl alcohol 8 in uivo and in vitro permits only the formation of (-)-secoisolariciresinol lb. The precise nature of the enzymatic process in this key coupling reaction is under active investigation.
To confirm and extend these radiochemical observations, we next undertook to demonstrate the conversion of [Ar-'H] secoisolariciresinol into [Ar-'Hlmatairesinol. Thus, (+)-[Ar-'Hlsecoisolariciresinols were prepared by deuterium exchange of aromatic protons of the unlabeled lignan with CF&02'H. The (?)-[Ar-"Hlsecoisolariciresinols, so obtained, were subjected to mass spectroscopic analysis. As can be seen (Fig.  4C), the parent molecular ion (M') for unlabeled secoisolariciresinol 1, previously noted at m/z 362 (Fig. 4B), was now shifted to an ion cluster centered at m/z 364, i.e. a partial aromatic substitution of H by D had occurred. This corresponds to the replacement of two to three aromatic hydrogens by deuterium. This observation was also confirmed by 'H NMR analysis. Following incubation of the (&)-[Ar-'H]secoisolariciresinols with the F. intermedia cell-free extract in the presence of NADP, the matairesinol 2b so obtained gave a cluster of ions now centered at m/z 360 (Fig. 40). This cluster is centered two to three mass units higher than that of natural abundance (-)-matairesinol2b (M', 358; Fig. 4E) indicating the presence of two to three deuterium atoms in the enzymatically formed (-)-matairesinol. Thus, the stereoselective conversion of (-)-secoisolariciresinol lb into (-)-matairesi-no1 2b had now been unequivocally demonstrated at the cellfree level.
In summary, we have detected enzymatic activities for lignan formation (Fig. 5), one of which is involved in the stereochemically controlled formation of (-)-secoisolaricire-sinol l b from coniferyl alcohol 8 and the other in the conversion of lignan l b to (-)-matairesinol 2b. More needs to be known about the coupling of the two phenylpropanoid units, in terms of how this enzyme (or enzymes) differ from typical peroxidase reactions. This is currently under investigation. Research directed to the elucidation of the biosynthetic pathways (intermediates and enzymes) involved in the formation of the more highly functionalized lignans, such as   , aniline (one drop) and piperidine (one drop) The reactlon mixture was stirred at 52°C for 21 h, then cooled (to ice-bath lemperature) and acidlfied to pH 2 with 2N HCI. To this solution was adued distllled water (ca. 10 ml). and the whole was extracted Wtth EtOAC (2 x 100 ml). The combined EtOAc solubles were washed successively with a saturaled NaHC03 solution until the washmgs became neutral, then with a Saturated NaCl soiution (ca. 10ml. hViCE), and dried (anhydrous Na2S04) The solvent was remnvsd in vacuo. and the product applled 10 a silica gel column (  (i)-Seco~soIar~ciresinois u: (A)-Mataireslnols (100 4 mg) were dtssoived ~n THF (9 mi, freshly distilled Over benzophenone and potasslum metal) under N2. The a sttrred suspenslon of LlAlH4 (88.6 mg) In dry THF (2 ml) Faliowlng stirrlng for an resulting salutlan was added dropwlse a1 mom temperature Over a perlod of 15 mln to addlllanal 1 h at the Same temperature. the reacllon mlxture was cooled to 0°C. Next, EtOAc (3 ml) was added dropwtse, and the whale was then poured onto dry Ice D~st~lled water (5 ml) was added to the resultlng suspenslon, wllh the organic solvent then removed tn vacuo. The sample was reconstituted In dlslilled water (10 ml) wlth the whole extracted With EtOAc (4 x 20 ml). The combined EtOAc solubles were washed wlth a saturated NaCl S O~U I I O~, dlled (anhydrous N a~S 0 4 ) and evaporated rn vacuo to yleld crude  WLb. (23.8 mg) were dissolved in CF3CO22H. generated from trifluoroacetc dlst\lled waler (1 ml) were added. and the resultlng solut?on was extracted wllh ElOAc (3 x 5 ml) The combined EtOAc solubles were washed wtth dlstiiled waler (1 mi). then wllh a Saturated NaCl Solution (2 x 1 ml). and dried (anhydrous NapSOq   (-)-secoisolariciresln~l lh was isolated by succewve purlflcatlon as before but with no addmon 01 unlabelled carrler. and subjected to mass spectroscopic analysis. Admmsfrafion of from I -0 to 20 mm and subpcled to analysts by hqwd Scmillation counting Next, aliquot$ (total 50 pi) were applied to the aforesacd column. and fractions correspondmg to matairesinol 2 were collecled, and sublected to chiral UPLC analysls Cell-free mtracf from F. ,ntsrmed,a Young shoots (5-10 cm long) Of F inlermwdia were exclsed by means of a razor, then washed with both tap and distilled water, and the leaves removed. The resulting slems (2.7 g fresh welghl) were cut Into small pieces by hand (sclssors), frozen (liq N2) and crushed in a mortar and pestle The powder so oblamed was further ground for 5-7 min with polyclar AT (0.54 9). acid-washed sea sand and 0.1 M polassium phosphate buffer (pH 7.0, 4 ml) contalnlng 10 mM dilhlothreitol The slurry was flllered through 4 layers of cheese-cloth, and the filtrate (3 rnl) cenlrtfuged (tS.000 x 9 , 20 rnfn). The resulting supernatant (2.7 ml) was again ffltered (Whalman GFA glass flbre filter), and an aliquot (1 5 ml) of the flllrate was applted to a Sephadex G-25 column [18 7 x 1 cm, Pharmacia, particle Size 50-150 pm (med~um)]. preequfllbrated in 0 t M potassium phosphate buffer (pH 7.0) contalnlng t o mM drth~othreilol. The fractlon excluded from the gel (1 5 ml) was collected and used as Ihe cell-free preparatlon.
Protein content 01 the preparation was 2.0 mg ml-1 on lhe basts 01 a BmRad Protein Assay usmg bovme serum albumin as standard I251