Evolution of Plant Defense Mechanisms

Pinoresinol-lariciresinol and isoflavone reductase classes are phylogenetically related, as is a third, the so-called “isoflavone reductase homologs.” This study establishes the first known catalytic function for the latter, as being able to engender the NADPH-dependent reduction of phenylcoumaran benzylic ethers. Accordingly, all three reductase classes are involved in the biosynthesis of important and related phenylpropanoid-derived plant defense compounds. In this investigation, the phenylcoumaran benzylic ether reductase from the gymnosperm, Pinus taeda, was cloned, with the recombinant protein heterologously expressed inEscherichia coli. The purified enzyme reduces the benzylic ether functionalities of both dehydrodiconiferyl alcohol and dihydrodehydrodiconiferyl alcohol, with a higher affinity for the former, as measured by apparent K m andV max values and observed kinetic3H-isotope effects. It abstracts the 4R-hydride of the required NADPH cofactor in a manner analogous to that of the pinoresinol-lariciresinol reductases and isoflavone reductases. A similar catalytic function was observed for the corresponding recombinant reductase whose gene was cloned from the angiosperm,Populus trichocarpa. Interestingly, both pinoresinol-lariciresinol reductases and isoflavone reductases catalyze enantiospecific conversions, whereas the phenylcoumaran benzylic ether reductase only shows regiospecific discrimination. A possible evolutionary relationship among the three reductase classes is proposed, based on the supposition that phenylcoumaran benzylic ether reductases represent the progenitors of pinoresinol-lariciresinol and isoflavone reductases.

Based on their high gene sequence homology, these enzymes appear to be phylogenetically linked (8). They also share comparable physiological roles for their products, in addition to having presumed similar catalytic modes of action, as shown in Fig. 1, A and B.
In recent years, a relatively large number of reductases with significant homology to both PLRs and IFRs (so-called IFR "homologs") have also been discovered (8), mainly through screening for genes induced by fungal elicitors (9) or in relation to other defensive responses to "stress" (10). To date, at least 19 such "homologs" have been reported in both woody and nonwoody angiosperms and gymnosperms, in plants as diverse as Arabidopsis thaliana (11), birch (Betula pendula), 2 Forsythia intermedia (8), maize (Zea mays) (10), pear (Pyrus communis), 3 poplar (Populus trichocarpa), 4 tobacco (Nicotiana tabacum) (9), and western red cedar (Thuja plicata) (12). Their roles in planta, however, have not been established; of those examined for enzymatic activity, none were able to catalyze either the reduction of isoflavones (9) or the reduction of pinoresinollariciresinol-type (8 -8Ј-linked) lignans (see below). Thus, their biochemical function has been unknown.
Lignans and isoflavonoids play important roles in plant defense. For example, their constitutive deposition significantly helps confer durability, longevity, and resistance to the heartwoods of many tree species against wood-rotting fungi. In this regard, western red cedar (T. plicata) heartwood formation results via deposition of ϳ20% of 8 -8Ј-linked lignans, such as plicatic acid 7 (15-23) (see Fig. 2), which help permit this species to grow to well over 60 meters tall and live for longer than 3000 years. The heartwoods of many Pinaceae species, such as loblolly pine (Pinus taeda), also contain significant levels of 8 -5Ј-linked (phenylcoumaran) lignans such as dehydrodiconiferyl alcohol 5 (see Fig. 1C), dihydrodehydrodiconiferyl alcohol 8 (Fig. 2), and related metabolites. These and their higher molecular weight (non-lignin) derivatives are laid down in the heartwood as protective postlignification infusions (4). Additionally, many tropical legume tree species contain significant levels of pterocarpan isoflavonoids, such as pterocarpin 9 in Pterocarpus indicus, thereby allowing these species to thrive in an environment highly conducive to fungal growth. Remarkably, almost nothing is known about the biochemistry of heartwood formation (4), despite the fact that it is responsible for ϳ90% of all woody biomass.
Lignan formation is not limited to significant deposition in heartwood and tissues attacked by fungi. For example, the 8 -5Ј-linked lignans, dehydrodiconiferyl alcohol triacetate 12 and tetrahydrodehydrodiconiferyl alcohol tetraacetate 13, ac-cumulate in small amounts in the leaves of Cryptomeria japonica (27). Moreover, dehydrodiconiferyl alcohol (DDC) 5 and its 7Ј-8Ј (allylic bond)-reduced derivative, dihydrodehydrodiconiferyl alcohol 8 (DDDC), appear to be ubiquitous throughout the plant kingdom, being found in plants as diverse as loblolly pine (P. taeda) (28), tobacco (29), ferns (30), and hornworts (31), i.e. spanning all major groups of vascular plants and suggesting the involvement of a universal defense system. Interestingly, the 8 -5Ј-linked lignans can also co-occur in many species with their (presumed) phenylcoumaran benzylic ether-reduced counterparts. As shown in Fig. 3, this includes examples such as dihydrodehydrodiconiferyl alcohol triacetate 12 with tetrahydrodehydrodiconiferyl alcohol tetraacetate 13 in C. japonica (27) and hancinone 14 with isodihydrofutoquinol A 15 in Piper wightii (32,33). In contrast to the 8 -8Ј-linked lignans and the isoflavonoids (1,5,8,24), however, there has been no explicit description of the biochemical pathways and enzymes involved in either the formation or subsequent metabolism of 8 -5Јlinked lignans.
From a biochemical perspective, enzymatic reduction of the benzylic ether of 8 -5Ј-linked lignans, such as DDC 5 and DDDC 8, might involve an analogous process to that envisaged for reduction of pinoresinol 1 (Fig. 1A) and isoflavonoids, such as 2Ј-hydroxypseudobaptigenin 3 (Fig. 1B), e.g. through formation and reduction of conjugated enone intermediates, as shown in Fig. 1C, or through reduction of some other enzymebound molecular species. Accordingly, we deduced that the PLR and IFR "homologs" may, in fact, be phenylcoumaran benzylic ether reductases (PCBERs). In order to test this hypothesis, and since suspension culture cells of the gymnosperm, loblolly pine, contain DDC 5, DDDC 8 (28), and related metabolites, we first constructed a cDNA library from RNA isolated from P. taeda suspension culture cells growing in 2,4-dichlorophenoxyacetic acid-containing medium (34). Next, when the F. intermedia pinoresinol-lariciresinol reductase cDNA was used as a probe, a "homolog" cDNA from loblolly pine was isolated, with its recombinant protein subsequently expressed in Escherichia coli. This protein, in the presence of NADPH, reduced the benzylic ether functionalities of both DDC 5 and DDDC 8. Another IFR "homolog" from the angiosperm, poplar (P. trichocarpa) 4 , when similarly expressed and tested for PCBER activity, also effected the same conversion. These findings thus identified a catalytic function for this class of enzymes as being capable of engendering phenylcoumaran benzylic ether reduction. This finding thus provided evidence that the so-called IFR "homologs" are indeed phylogenetically linked to PLR and IFR and demonstrated further the evolution and conservation of the phenylpropanoid pathway throughout the plant kingdom (8). In contrast to the enantiospecific PLR and IFR, however, the phenylcoumaran benzylic ether reductases only displayed regiochemical specificities, since both (ϩ)-and (Ϫ)-enantiomers of DDC 5 were effectively reduced.

EXPERIMENTAL PROCEDURES
Plant Materials-P. taeda cell suspension cultures were maintained as described previously (34), in Brown and Lawrence medium containing 2,4-dichlorophenoxyacetic acid. Cells were harvested by filtration 7 days after transfer to fresh media, frozen in liquid nitrogen, and stored at Ϫ80°C. Populus deltoides ϫ P. trichocarpa, clone 064 (Afocel, France) and P. trichocarpa cv. Trichobel were grown in the Institute of Forestry and Game Management nursery (Geraardsbergen, Belgium).
General Methods-All molecular biological techniques, unless expressly described otherwise below, were performed according to standard methods (35,36). Protein purification procedures using a fast pro-tein liquid chromatography (Amersham Pharmacia Biotech) system were carried out at 4°C, whereas those employing the BioCad (PerSeptive Biosystems) system were conducted at ambient temperature. All chromatographic eluents were monitored at 280 nm. Protein concentrations were determined with the Bio-Rad protein determination kit. Polyacrylamide gel electrophoresis used gradient (4 -15% linear gradient; Bio-Rad) gels under denaturing and reducing conditions in the Laemmli buffer system (37), followed by visualization of the proteins by silver staining (38).
Other Materials-All solvents and chemicals used were reagent or high performance liquid chromatography (HPLC) grade. Taq thermostable DNA polymerase, competent NovaBlue cells, and radiolabeled nucleotide ([␣-32 P]dCTP) were purchased from Promega, Novagen, and NEN Life Science Products, respectively. The pCRII TA cloning kit was from Invitrogen. Restriction endonuclease NdeI was from New England Biolabs.
Oligonucleotide primers for polymerase chain reaction (PCR) and sequencing were obtained from Life Technologies, Inc. Geneclean II ® kits for PCR fragment purification were from BIO 101 Inc., with the gel-purified DNA concentrations determined by comparison with a low DNA mass ladder (Life Technologies, Inc.) in 1.3% agarose gels.
Instrumentation-A Lambda 6 UV-visible spectrophotometer (Perkin-Elmer) was used for recording all RNA and DNA determinations at A 260 , and a Temptronic II thermocycler (Thermolyne) was employed for all PCR amplifications. The Wizard ® Plus SV Minipreps DNA Purification System (Promega) was used to purify plasmid DNA for sequencing using an Applied Biosystems model 373A automated sequencer. All HPLC separations were performed on either a Millennium TM (Waters Inc.) or Alliance TM (Waters Inc.) instrument. Amino-terminal protein sequencing was carried out as described previously (1). Matrix-assisted laser desorption ionization-time of flight mass spectrometry was performed on a VG 7070 at the Laboratory for Bioanalysis and Biotechnology at Washington State University.
P. taeda cDNA Library Synthesis-Total RNA (100 g/g, fresh weight) was obtained using the method of Dong and Dunstan (39) from frozen loblolly pine (P. taeda) cells grown as described above. A P. taeda cDNA library was constructed using 5 g of purified poly(A) ϩ mRNA (Oligotex-dT TM suspension; Qiagen) with the ZAP-cDNA ® synthesis kit, the Uni ZAP TM XR vector, and the Gigapack ® II Gold packaging extract (Stratagene), with a titer of 1 ϫ 10 6 plaque-forming units for the primary library. The amplified library (1 ϫ 10 9 plaque-forming units/ ml; 120 ml total) was used for screening (1).
DNA Probe Synthesis-The 5Ј-end of the previously isolated pinoresinol-lariciresinol reductase cDNA (PLR-Fi1) was used to construct a probe, as described previously (1), to screen the P. taeda cDNA library for similar/homologous genes.
Library Screening-600,000 plaque-forming units of the P. taeda amplified cDNA library were screened for genes homologous to PLR-Fi1. This was done exactly as described by Dinkova-Kostova et al. (1), with the exception that hybridization temperatures were reduced to 45°C. 20 positive plaques were purified through two more rounds of screening.
In Vivo Excision and Sequencing of Putative Reductase cDNAcontaining Phagemids-Purified cDNA clones were rescued from the phage following Stratagene's in vivo excision protocol. Both strands of several different cDNAs that coded for genes homologous to PLR-Fi1 were completely sequenced using overlapping sequencing primers.
Sequence Analysis-DNA and amino acid sequence analyses were performed using the Unix-based GCG Wisconsin Package (40,41) and the ExPASy World Wide Web molecular biology server (42).

Expression in Escherichia coli as a Fusion Protein-
The open reading frame of the putative reductase from P. taeda was in frame with the ␤-galactosidase gene ␣-complementation particle in pBluescript. Thus, its recombinant fusion protein was expressed in NovaBlue cells as described previously (1) and then assayed for both phenylcoumaran benzylic ether reductase and pinoresinol-lariciresinol reductase activities, as described below.
Transfer of the Loblolly Pine (P. taeda) Phenylcoumaran Benzylic Ether Reductase cDNA into pSBETa-The overexpression plasmid, pS-BETa (43), which contains the NdeI restriction site for native expression, was used. Since the putative reductase from P. taeda had no internal NdeI sites, two primers were designed to introduce NdeI sites at the start methionine (primer PT-ATG-NdeI, TTC AGG GCC CAT ATG GGA AGC AGG AGC AGG ATA CTC) and in the 3Ј-end untranslated region (primer PT-Rev-NdeI, TGT CGA ATA CAT ATG AAA GGC GAT AAC CAA CAA TTT). These two primers (5 pmol each) were used in five PCRs with 10 ng of the cDNA encoding the P. taeda putative reductase as described (1) and subcloned into the pCRII plasmid according to Invitrogen's instructions. The resulting pCRII plasmid containing the putative reductase was purified and digested with NdeI, with the resulting ϳ1-kilobase pair fragment gel purified and ligated into pSBETa previously digested with NdeI and transformed into competent NovaBlue cells. The resulting pSBET construct containing the putative phenylcoumaran benzylic ether reductase was purified, and the expression region, containing the desired cDNA, was sequenced completely on both strands to verify that no mutations had been introduced during PCR (see Fig. 4). The amino acid sequence of the protein was then compared with PLR and IFR amino acid sequences (see Fig.  5).
Overexpression of the Native Enzyme in E. coli-The resulting pS-BETa plasmid containing the putative reductase from P. taeda was transformed into B834(DE3) E. coli cells for expression. High level expression of the putative reductase was achieved by inoculating 1 liter of LB broth, supplemented with 50 g/ml kanamycin, with 2-4 ml of an overnight grown 10-ml culture in the same medium. The cells were then allowed to grow at 37°C with shaking at 250 rpm until a density of A 600 ϭ 0.65 was reached, at which point the growth conditions were changed to 20°C with shaking at 265 rpm. Production of the reductase was induced by the addition of isopropyl ␤-D-thiogalactopyranoside to 1 mM final concentration, once the cells had reached a temperature of ϳ22°C (about 30 min after the incubator temperature was turned down). The cells were then allowed to grow for 21 h before harvest by centrifugation for 25 min at 3000 ϫ g in 4 ϫ 250-ml centrifuge bottles, after which the pellets were frozen for at least 2 h at Ϫ80°C to aid in cell lysis.
Crude Protein Preparation-The four pellets were thawed at room temperature and resuspended in 2 ϫ 10 ml each of Buffer A (20 mM Tris⅐HCl, pH 8.0; 2 mM EDTA; 1 mM phenylmethylsulfonyl fluoride; 5 mM DTT) and then combined and sonicated for 3 ϫ 30 s. The sonicates were centrifuged for 30 min at 20,000 ϫ g to pellet cellular debris and filtered through a 0.2-m syringe filter. The resulting filtrate (290 ml) was subjected to ammonium sulfate precipitation, with the 40 -70% ammonium sulfate saturation fraction, after centrifugation for 30 min at 20,000 ϫ g, desalted back into Buffer A over PD-10 columns (Amersham Pharmacia Biotech).
Affinity Column (Affi-Blue Gel) Purification-A 1.6-cm diameter, 11.5-cm-long (23-ml bed volume) Affi-Gel blue gel (Bio-Rad) column was preequilibrated in Buffer A. The desalted 40 -70% fraction (15 ml, 513 mg) was then applied to the column. After washing with 300 ml of Buffer A (1 ml/min), the reductase was eluted by running a linear gradient from 0 to 100% Buffer B (Buffer A plus 5 M NaCl) in 30 ml, holding at 100% Buffer B for 60 ml and returning from 100% Buffer B to 100% Buffer A in 30 ml. 4-ml fractions were collected, with the reductase being eluted in fractions 33-43. These fractions were then pooled, concentrated to ϳ5 mg/ml in a Centricon 10 microconcentrator (Amicon, Inc.), desalted over PD10 columns, and assayed for activity.
Anion Exchange Chromatography-The resulting enzyme solution (118 mg) was next applied to a POROS 20 QE Perfusion anion exchange column, preequilibrated in Buffer C (50 mM Bis-Tris propane, pH 6.8, 5 mM DTT). The desired protein did not bind to the column, whereas the majority of the contaminating E. coli proteins remained bound under these conditions.
Cation Exchange Chromatography-The resulting enzyme solution (59 mg) was next applied to a POROS 20 SP Perfusion cation exchange column, preequilibrated in Buffer C. Again, the desired protein did not bind to the column, whereas the rest of the contaminating E. coli proteins remained bound. 36 mg of purified reductase protein were thus obtained, this being apparently homogeneous, as revealed by SDSpolyacrylamide gel electrophoresis with visualization by silver staining (see Fig. 6 and "Results and Discussion").
Cloning of P. trichocarpa cv. Trichobel Benzylic Ether Reductase-Wood of P. deltoides ϫ P. trichocarpa, clone 064, was sampled in May 1992 by removal of a 2-year-old branch from a 20-year-old tree and separating the bark from the wood. Wood pieces were immediately frozen in liquid nitrogen and stored at Ϫ70°C. These were subsequently used for two-dimensional gel electrophoresis and subsequent microsequencing of the most abundant protein in the gels. 5 For the sampling of P. trichocarpa cv. Trichobel, 3-year-old trees were harvested, with the bark again separated, and the selected woody tissues immediately frozen in liquid nitrogen and stored at Ϫ70°C. Woody tissue (ϳ10 cm long) was obtained from the stem 10 cm above ground level. A root sample was also harvested at ϳ5 cm below ground level. Peptide sequences obtained from the most abundant protein in twodimensional protein gels of the wood sampled from P. deltoides ϫ P. trichocarpa, clone 064, were as follows: 1) TGYIGK, 2) PTFALV, 3) GVTLIHGDV, 4) FFPSEFGM, 5) TFVPETK, and 6) TTVEEYLDQFV. 5 The amino acid sequence of two internal peptides (numbers 3 and 4 above) and one C-terminal peptide (number 6) were next used to design degenerate oligonucleotide primers: PT primer 2 (GGI GTI ACI (C/T)TC (C/T)TC IAC IGT IGT (G/A)TA)), respectively. Purified DNA from P. deltoides ϫ P. trichocarpa, clone 064, leaf cDNA library (44) was then used as the template in PCRs, with PT primer 2 and either PT primer 5 or PT primer 9. The PCR conditions were set at 30 cycles of 94°C for 1 min, 50°C for 1 min, and 72°C for 1 min, with an initial denaturation step of 94°C for 4 min and a final step of 72°C for 5 min. DNA blot analysis of the PCR products using a 32 P-labeled PIFR4 (7) as a probe revealed two hybridizing bands: a 160-base pair fragment for primer combination 2/5 and a 750-base pair fragment for primer combination 2/9. The larger fragment was next cloned in a T-vector (Amersham Pharmacia Biotech) and used as a 32 P-labeled probe to screen a P. trichocarpa cv. Trichobel xylem cDNA library (obtained as a ZAPII cDNA library from xylem harvested in May 1992), according to the manufacturer's instructions (Stratagene). The first round of screening resulted in a high number of positive plaques, with 20 being subjected to two additional rounds of screening. Eleven positive clones were purified; in vivo excision was performed according to the manufacturer's instructions (Stratagene); and four full-length cDNAs, present in the EcoRI-XhoI site of Bluescript II SK(Ϫ), were obtained, sequenced, and found to differ only in length. The longest, called PCBER-Pop1, was used for further analysis.
Expression of the Poplar (P. trichocarpa) Reductase in E. coli-To assess whether other IFR "homolog" proteins were also able to reduce DDC 5 or DDDC 8, the angiosperm IFR "homolog" recombinant reductase from P. trichocarpa, PCBER-Pop1, was analyzed for reductase activity. A pGEX-2T-based plasmid, containing the cDNA in frame with Schistosoma japonicum glutathione S-transferase, was constructed according to the manufacturer's instructions (Amersham Pharmacia Biotech). This expression plasmid was then transformed into NovaBlue E. coli cells, which were then grown, induced for protein production as described above for the pSBETa-based system, except that carbenicillin (50 g/ml) was used as antibiotic instead of kanamycin; and finally pelleted (3000 ϫ g, 25 min) for storage at Ϫ80°C. After cell lysis in phosphate-buffered saline (140 mM NaCl, 2.7 mM KCl, 10 mM K 2 HPO 4 , 1.8 mM KH 2 PO 4 ) containing 5 mM DTT, the protein was purified over glutathione-Sepharose 4B resin according to Pharmacia's instructions. The N-terminal glutathione S-transferase fusion was then removed from 50 mg of purified protein by cleavage at 4°C for 16 h with 500 units of thrombin (Amersham Pharmacia Biotech), with the resulting protein tested for both phenylcoumaran benzylic ether reductase and PLR activities, as described below.
Chemical Synthesis of (Ϯ)-Dehydrodiconiferyl Alcohol (5)-To a solution of E-coniferyl alcohol 16 (400 mg, 2.2 mmol) in acetone (15 ml) was added FeCl 3 ⅐6H 2 O (1.5 g, 5.5 mmol) in H 2 O (35 ml), with the resulting suspension stirred for 10 min at ambient temperature and then extracted with ethyl acetate (3 ϫ 50 ml). The ethyl acetate solubles were combined, washed with H 2 O (2 ϫ 30 ml), evaporated to dryness in vacuo, and reconstituted in chloroform (10 ml). This was then applied to a silica gel column (3 ϫ 20 cm) eluted with a gradient solvent system from chloroform to chloroform/acetone (3:2). Fractions containing the desired product were combined and evaporated to dryness in vacuo, recrystallization of which from chloroform gave DDC 5 (78 mg, 0.2 mmol, 9% yield) as a white powder. 1  Chemical Synthesis of (Ϯ)-Dihydrodehydrodiconiferyl Alcohol (8)-This procedure was based on a previously reported approach (28). To a solution of 5 (100 mg, 0.28 mmol) in ethyl acetate (5 ml) was added palladium-charcoal (10%, 50 mg), and the resulting solution was stirred under an H 2 atmosphere for 10 min at ambient temperature. The catalyst was removed by filtration, washed with ethyl acetate (10 ml), with the ethyl acetate solubles combined and evaporated to dryness in vacuo. The resulting residue was reconstituted in chloroform and then applied to a silica gel column (1 ϫ 10 cm) eluted with chloroform/ acetone (4:1). The fractions containing the desired product were combined and evaporated to dryness in vacuo to afford DDDC 8 (80 mg, 0.22 mmol, 79%). 1  Chemical Synthesis of (Ϯ)-Tetrahydrodehydrodiconiferyl Alcohol (17)-To a solution of 5 (100 mg, 0.28 mmol) in methanol (5 ml) was added palladium-charcoal (10%, 50 mg), and the resulting solution was stirred under an H 2 atmosphere for 8 h at ambient temperature. After the catalyst was removed by filtration, the resulting solution was washed with methanol (10 ml), with the methanol solubles combined and evaporated to dryness in vacuo. The resulting residue was reconstituted in chloroform and then applied to a silica gel column (1 ϫ 10 cm) eluted with chloroform/acetone (4:1). The fractions containing the desired product were combined and evaporated to dryness in vacuo to afford tetrahydrodehydrodiconiferyl alcohol (TDDC) 17 (60 mg, 0.17 mmol, 61%). 1  Enzymatic Formation of (Ϯ)-Tetrahydrodehydrodiconiferyl Alcohol (17)-30 assays containing 2.5 mM DDDC 8 each were conducted exactly as above to give, following extraction and purification as before, TDDC 17 (380 g, 1.1 mol, 9.9%).
Reversed Phase HPLC Separation of Dehydrodiconiferyl Alcohol 5 and Analogs 6, 8, and 17-Separation of dehydrodiconiferyl alcohol 5, dihydrodehydrodiconiferyl alcohol 8, isodihydrodehydrodiconiferyl alcohol 6, and tetrahydrodehydrodiconiferyl alcohol 17 was accomplished using a reversed-phase column (Symmetry Shield RP 8 , 3.9 ϫ 150 mm, Waters) eluted with an acetonitrile, 3% acetic acid (in H 2 O) gradient solvent system at a flow rate of 1 ml/min with detection at 280 nm. The column was pre-equilibrated in acetonitrile, 3% acetic acid in H 2 O (1:9), this being held for 5 min following sample introduction. A linear gradient to 1:3 (acetonitrile, 3% acetic acid in H 2 O) was then applied over 30 min, followed by a linear gradient to 100% acetonitrile over 25 min.  (1). After a 6-h incubation at 22°C, the assay mixture was extracted with ethyl acetate (2 ϫ 500 l). The ethyl acetate-soluble fraction was evaporated to dryness in vacuo and then reconstituted in 100 l of acetonitrile, 3% acetic acid in H 2 O (1:9) and subjected to reversed-phase HPLC as described above, with both UV and radiochemical detection. 1-ml fractions were collected, with an aliquot (100 l) of each removed for scintillation counting to determine the level of incorporation of 3 H into the assay products. Controls using denatured enzyme (5 min, 100°C) or no (Ϯ)-dehydrodiconiferyl alcohol 5 substrates were also performed.

Chiral Column HPLC Separation of the (ϩ)-and (Ϫ)-Enantiomers of
Nonradioactive Assays for Dehydrodiconiferyl Alcohol Reductase Activity-Each 150-l assay consisted of 22 mM MES:Bis-Tris propane, pH 6.5, 20 l of the protein solution at the corresponding stage of purity, 5 mM DTT, 2 mM (Ϯ)-dehydrodiconiferyl alcohols 5, and 4 mM NADPH. After a 3-h incubation at 30°C, the assay mixture was boiled for 3 min and centrifuged (17,000 ϫ g, 3 min), with an aliquot (125 l) subjected to reversed-phase HPLC, after 16.6 l of acetonitrile was added. Quantification of formation of isodihydrodehydrodiconiferyl alcohol 6 was measured using a standard curve. Controls using denatured enzyme (5 min, 100°C) or no NADPH cofactor were also performed.
Enzyme Characterization-Optimum temperature and pH were determined using standard (nonradioactive) assay conditions as described above, except that the buffer concentration was changed to 19 mM and the protein following the Affi-Blue column chromatographic step was used (30 l). For temperature optimum, incubations were performed at constant pH (7.0) with varying temperature (6 -58°C); for pH optimum, incubations were performed at constant temperature (30°C) with varying pH (5.5-9.5). Initial velocity kinetics were analyzed by assaying the protein activity under standard (nonradioactive) conditions at pH 7.0,

FIG. 6. SDS-polyacrylamide gel electrophoresis showing the steps involved in the purification of the P. taeda phenylcoumaran benzylic ether reductase from the induced E. coli cell-free extracts (lane 1, lysate) to the purified protein (lane 5, cation exchange).
but with 11 different concentrations of lignan substrate (0.167-2.5 mM) and at 22°C for 6 h, while holding the NADPH concentration constant at 5 mM. To determine whether the 4R-or the 4S-hydride of the NADPH cofactor was utilized in the reduction catalyzed by the enzyme, radiochemical assays were performed as described above with specifically labeled [4R-3 H]NADPH (1.67 mM, 10.7 MBq/mmol) or [4S-3 H]NADPH (1.67 mM, 9.8 MBq/mmol) (1) and analyzed for radiochemical incorporation into the isodihydrodehydrodiconiferyl alcohol 6 product.
Chiral HPLC Analysis of IDDDC 6 Generated by Action of Phenylcoumaran Benzylic Ether Reductase-Assays using DDC 5 as substrate were performed as above. The remaining substrate 5 and resulting product IDDDC 6 were separated by Symmetry Shield HPLC, collected individually, dried in vacuo, redissolved in methanol, and subjected to chiral HPLC analysis as described above. Both (ϩ)-and (Ϫ)-enantiomers of substrate DDC 5 were detected in equal amount as were both enantiomers of product IDDDC 6.
Radiochemical Assays for Pinoresinol-Lariciresinol Reductase Activity-Pinoresinol reductase and lariciresinol reductase activities were evaluated by monitoring formation of 3 H-labeled product as described elsewhere (12).
Nonradioactive Assays for Pinoresinol-Lariciresinol Reductase Activity-Pinoresinol and lariciresinol reductase activities were further evaluated in assays performed as described in Ref. 12  pinoresinols 1 or (Ϯ)-lariciresinols 2 were used as cofactors and substrates, respectively; and no radiochemical carriers were added. Enzymatic reactions were stopped by boiling for 3 min, with 50 l of methanol then added. After assay mixtures were centrifuged (17,000 ϫ g for 3 min), 150 l of the resulting mixture was subjected to reversed phase HPLC (12) to determine whether either pinoresinol 1 or lariciresinol 2 had been reduced.

RESULTS AND DISCUSSION
In recent years, a relatively large family of three classes of reductases (8) has been identified from diverse members of the angiosperms and gymnosperms, with each correlated to functions associated with plant defense. Two of these classes were demonstrated to play significant roles in the biosynthesis of important plant defense phenylpropanoid-derived compounds, based on known catalytic function, i.e. the PLRs and the IFRs (8). Members of the third class, however, had no known function but were described as IFR/PLR "homologs," since they share a significant level of homology, with ϳ60 -70% similarity at the amino acid level to IFRs and PLRs. All three classes possess NADPH binding sites and appear to be cytosolic enzymes, since they lack secretory pathway signal sequences. Additionally, they may be regulated in a comparable manner, not only at the transcriptional level (24) but also at the enzymatic level by protein phosphorylation, as evidenced by several potential phosphorylation sites that are conserved among all of the members (1,8).
The PLR/IFR "homologs," on the other hand, had no known catalytic function, since of those examined, none were able to catalyze either the reduction of isoflavonoids (9) or 8 -8Ј-linked lignans (see below), and no other role has been forthcoming. However, their formation is inducible, both in cell culture and in planta, by factors such as sulfur starvation (oxidative stress) (10), fungal elicitors (9), and UV irradiation (48), indicative of a general plant defense role. DDC 5, DDDC 8, and analogs thereof are 8 -5Ј-linked (phenylcoumaran) lignans that appear to be ubiquitous in the plant kingdom, being found in a wide array of plant families, such as, for example, the Asteraceae (49), Pinaceae (50), and Urticaceae (51). Although they were originally proposed to be formed either as intermediates in lignification (52)(53)(54) and/or in their glycoside forms as plant growth regulators (29,55,56), no consideration was ever given, from a biochemical perspective, to their further metabolism.
It could, therefore, be considered that both DDC 5 and DDDC 8 might undergo comparable reductive transformations, analogous to that catalyzed by PLR and IFR, and hence provide an identification of the biochemical role of the so-called "homologs." Cloning, Expression, and Purification of PCBER-Pt1-Given the above findings, we investigated whether the "homolog" reductases catalyze reduction of DDC 5 and/or DDDC 8. Of particular interest was whether these reductions, as for their PLR and IFR counterparts, would display any enantiospecific preference or whether they would only be regiospecific. In the approach undertaken in this study, a -phage based cDNA library was first constructed from RNA isolated from P. taeda cell suspension cultures. These cells had previously been shown to produce DDC 5, DDDC 8 (28), and related lignan metabolites upon treatment with 8% sucrose, 20 mM potassium iodide. This cDNA library was then screened for potential reductases using the 5Ј-end (ϳ400 base pairs) of the F. intermedia pinoresinollariciresinol reductase (PLR-Fi1) cDNA (1) as a probe (this approach had previously been successful in screening F. intermedia, which resulted in identification of two "homologs" (1,8)). Accordingly, moderate stringency conditions identified many potentially homologous cDNAs in the P. taeda library, 20 of which were purified through two additional rounds of screening. The pBluescript-based phagemids that carried the cDNAs of interest were excised from the -phage and ultimately gave only a single gene with homology to the F. intermedia pinoresinol-lariciresinol reductase from the P. taeda cell suspension culture cDNA library.
Analysis of its cDNA sequence revealed that it had ϳ66% similarity and ϳ45% identity to PLR from F. intermedia, and thus fell into the category of the IFR/PLR "homologs" (8). As shown in Fig. 4, this P. taeda reductase cDNA encodes a 308amino acid protein, corresponding to 33.6 kDa. It also appears to be cytosolic, possessing no signal sequences for targeting to the secretory pathway, mitochondria, chloroplasts, or peroxisomes. Additionally, the protein contains the conserved GX(X)GXXG consensus sequence, indicative of an NAD(P)Hbinding domain, as well as 12 potential phosphorylation sites (see Fig. 4), most of which are conserved in PLR, IFR, and the IFR "homologs" as well (8). The high similarity between these classes of reductases is further illustrated in Fig. 5, which compares the amino acid sequence of the P. taeda reductase (PCBER-Pt1) with several PLR, IFR, and "homolog" protein sequences. The presumed significance of this is discussed in detail below.
To first obtain catalytically active protein, native expression was performed in E. coli utilizing the overexpression plasmid, pSBETa (43), which contains the pET3a expression cassette, with an NdeI restriction site available for subcloning to yield native expression, as well as the argU gene (for production of the rare AGA-Arg tRNA 4 Arg ) and kanamycin resistance (for high plasmid stability in liquid cultures). Since the putative reductase from P. taeda had no internal NdeI sites, subcloning into pSBETa was relatively straightforward. Two primers were used in PCRs to introduce NdeI sites at the start methionine (primer PT-ATG-NdeI) and in the 3Ј-end untranslated region (primer PT-Rev-NdeI) of the putative reductase. The regions of the cDNA sequence used for primer design are underlined in Fig. 4. The resulting PCR fragment was subcloned into a PCR cloning plasmid (pCRII; Invitrogen) and excised by cleavage with NdeI. The ϳ1-kilobase pair fragment containing the putative reductase was gel-purified and ligated into pSBETa, which had been previously digested with NdeI. The resulting pSBET construct containing the putative reductase was transformed into competent E. coli cells and purified, and the expression region, containing the desired cDNA, was sequenced completely on both strands to verify that no mutations had been introduced during PCR.
Induction of the reductase in the pSBET overexpression system was accomplished by transforming B834(DE3) expression cells with the pSBET construct and then growing at 37°C four 1-liter cultures in LB broth supplemented with 50 g/ml kanamycin until they reached a density of A 600 ϳ0.6, at which time they were cooled to ϳ22°C and inoculated with isopropyl ␤-Dthiogalactopyranoside to a final concentration of 1 mM. After growth for 21 h at ϳ20 -22°C, the cells were harvested by centrifugation.
A 12-fold purification to apparent homogeneity of the heterologously expressed native enzyme was accomplished in four chromatographic steps following cell lysis and cellular debris removal by centrifugation (see "Experimental Procedures" and Table I). The first step employed ammonium sulfate precipitation, where the desired protein precipitated in the 40 -70% saturation fraction. This fraction of the protein (513 mg) was desalted into Buffer A and applied to an affinity (Affi-Blue gel) column and eluted with a linear NaCl gradient. The resulting protein (118 mg) was again desalted into Buffer A and then subjected to consecutive anion (POROS 20 QE) and cation (POROS 20 SP) exchange chromatography. Both chromatographic steps were conducted under conditions (pH 6.8) where the desired protein did not bind, whereas the contaminating E. coli proteins did and were thus removed. Fig. 6 shows an SDS-polyacrylamide gel of the reductase at each step of the purification scheme. After the final purification step, the protein was subjected to SDS-polyacrylamide gel electrophoresis and blotted to a polyvinylidene fluoride membrane, and the amino terminus thereof was sequenced. The amino acid sequence thus obtained matched exactly that deduced from the P. taeda cDNA sequence. The purified protein was also subjected to matrix-assisted laser desorption ionization-time of flight mass spectrometry, which showed a molecular mass of 33.6 kDa for the recombinant protein. These results indicated that the target reductase had been expressed.
Enzymatic Activity Determination-We next evaluated the P. taeda recombinant reductase for either its ability to convert the phenylcoumaran lignans, DDC 5 and/or DDDC 8, into the benzylic ether-reduced lignans, IDDDC 6 and TDDC 17, respectively, or to reduce the lignans, pinoresinol 1 and lariciresinol 2.
In order to do this, the potential substrates and products were first obtained (semi)synthetically, i.e. racemic (Ϯ)-dehydrodiconiferyl alcohols 5 were synthesized via FeCl 3 oxidation of E-coniferyl alcohol 16. DDC 5 was then subjected to catalytic hydrogenation (10% palladium-charcoal, H 2 , 10 min) to produce the corresponding allylic bond-reduced derivative (Ϯ)-DDDC 8 in 79% yield, whose mass and 1 H/ 13 C NMR spectra revealed that only reduction of the allylic 7Ј-8Ј-bond had occurred. In an analogous manner, further reduction (10% palladium-charcoal, H 2 , 8 h) of DDC 5 produced (Ϯ)-TDDC 17 in 61% yield. Its EIMS gave a molecular ion [M ϩ ] at m/z 362 and expected fragments at m/z 208 and 137, whereas its 1 H NMR spectrum revealed the presence of five methylene groups (␦ 1.79, 2.57, 2.91, 3.57, and 3.81). On the other hand, IDDDC 6 was obtained semisynthetically, since it was generated by the action of the P. taeda reductase on DDDC 8. Its EIMS displayed a molecular ion [M ϩ ] at m/z 360, with main fragment ions at m/z 206 and 137. Its 1 H NMR spectrum displayed characteristic olefinic resonances at ␦ 6.20 and 6.52, together with three methylenic functionalities at ␦ 2.95, 3.85, and 4.30.
With all potential substrates and products obtained in racemic form, conditions were next identified for their facile reversed-phase HPLC separation. As shown in Fig. 7A, a near base-line separation of all four lignans was obtained (see "Experimental Procedures" for elution conditions). Additionally, each lignan was subjected to chiral column HPLC analysis using a Daicel Chiralcel OF column, employing hexane/2-propanol isocratic solvent systems (see "Experimental Procedures"), which permitted determination of their enantiomeric compositions.
Assays for DDC 5 and/or DDDC 8 reductase activities were next performed (see Fig. 7, B, C, E, F), as described under "Experimental Procedures." As shown in Fig. 7, B and C, the P. taeda reductase converted both DDC 5 and DDDC 8 into ID-DDC 6 and TDDC 17, respectively. In each case, identification of the products was determined using HPLC retention volumes, UV and 1 H NMR spectroscopy, and liquid chromatography-mass spectrometry fragmentation analysis. For illustrative purposes, the liquid chromatography-mass spectrometry spectra of the enzymatic products IDDDC 6 and TDDC 17 are shown in Fig. 7, E and F, respectively. Using DDC 5 as a substrate (Fig. 7D), the enzymatically generated IDDDC 6 ( , and 137 (benzylic fragment). Control assays with either DDC 5 or DDDC 8, conducted in the absence of NADPH or with boiled (denatured) enzyme, showed no conversion into IDDDC 6 or TDDC 17, respectively. Thus, the so-called IFR/ PLR "homologs" were able to function as phenylcoumaran benzylic ether reductases (PCBERs; see Fig. 1C). Fig. 8 shows the proposed biosynthetic pathway, based on these results, to the 8 -5Ј-linked lignans in plants such as P. taeda.
As shown in Table II, further characterization of the P. taeda PCBER-Pt1, using DDC 5 as substrate, gave a pH optimum plateau from 6.4 to 7.0 and a temperature optimum at 49°C. In addition, it is also a type A reductase, since transfer of the 3 (Table II). This observation is in agreement with that noted for both PLR (1) and IFR (47).
Initial velocity studies were next carried out in order to determine whether PCBER-Pt1 preferred one substrate over the other (DDC 5 versus DDDC 8). As shown in Table II, the specific activity (V max ) is 2-fold higher for DDC 5 than for DDDC 8, whereas the K m for DDC 5 reduction is 3 times lower than that for DDDC 8 reduction, i.e. suggesting a slightly  (13), in contrast to that observed for DDC 5. This is also supported by the increase in K m for DDDC 8 reduction, indicating that its binding is less efficient than for DDC 5. These results, therefore, demonstrate that the P. taeda PCBER-Pt1 has a higher affinity for DDC 5 as substrate, suggesting that the biosynthesis of molecules such as tetrahydrodehydrodiconiferyl alcohol tetraacetate 13, in C. japonica, may involve initial formation of the benzylic ether-reduced intermediate prior to allylic bond reduction. It was next important to ascertain whether the angiosperm "homologs" possessed PCBER activity. Accordingly, we obtained the cDNA encoding an IFR "homolog" from P. trichocarpa, PCBER-Pop1, in a pGEX-based system (see "Experi-mental Procedures"), with the cDNA in frame with N-terminal glutathione S-transferase. After expression in E. coli, purification of the protein over glutathione-Sepharose 4B (see "Experimental Procedures"), and removal of the glutathione S-transferase fusion by thrombin cleavage, the recombinant protein was assayed for PCBER activity. This protein was able to reduce the benzylic ether bond of DDC 5, with a specific activity (53 nmol/mg/h) comparable with the reduction by PCBER-Pt1 of DDDC 8 (see Table II). Thus, the angiosperm IFR "homologs" can function as phenylcoumaran benzylic ether reductases as well.
To address the issue of whether PCBER-Pt1 displayed any enantiospecific effect toward its substrates, comparable with both PLR (1,12) and IFR (45)(46)(47), the products of enzymatic conversions of DDC 5 and DDDC 8 were evaluated by chiral HPLC analysis (see "Experimental Procedures"). Unlike the other members of this large family of reductases, however, PCBER-Pt1 showed no apparent enantiospecificity toward either substrate, with racemic products being formed (data not shown).
Studies were also conducted to establish whether pinoresinol 1 or lariciresinol 2 (8 -8Ј-linked lignans) could serve as substrates for PCBERs (see "Experimental Procedures"), using PCBER-Pop1 and PCBER-Pt1 (with the native expressed recombinant enzyme or with the ␤-galactosidase fusion protein (1)) as before. It was found, however, that with assays utilizing pinoresinol 1 or lariciresinol 2 as substrate and either [4R-3 H]NADPH or unlabeled NADPH as cofactor, no product formation was observed. This was still the case, even when assays were extended for up to 24 h (i.e. under conditions that would have depleted all available substrate with the F. intermedia reductase (PLR-Fi1) (1)). In addition, as discussed earlier, the IFR "homologs" are apparently unable to reduce isoflavones (9). Accordingly, the only catalytic function identified to date for both PCBER-Pt1 and PCBER-Pop1 involves that of the first known example of phenylcoumaran benzylic ether cleavage of the 8 -5Ј-linked lignans DDC 5 and DDDC 8. Since recent work with pinoresinol-lariciresinol reductases from western red cedar has demonstrated an apparent K m of ϳ0.62 mM for pinores-  inol 1 as substrate, 6 a value very similar to that obtained for PCBER-Pt1 with DDC 5, it seems likely that the natural substrate for the latter enzyme is either DDC 5 or some molecule that is very similar. The specific activities of PLRs from both western red cedar and Forsythia intermedia, 6 at ϳ25,000 nmol/ mg/h, however, are about 185 times greater than that observed for PCBER-Pt1. This discrepancy in reaction rate cannot be explained at the present time, but it may represent differential regulation of these two enzyme classes, where PCBERs may require external factors for activation, such as phosphorylation or allosteric affectors. To compare the relationships between these phenylcoumaran benzylic ether reductases with PLRs, IFRs, and other "homologs," a dendogram was produced, as shown in Fig. 9, based on sequence similarity at the amino acid level. As can be seen in cluster 1, all of the PLRs group together, along with a "homolog" from Lupinus albus (14), which may also encode a PLR. The IFRs, together with the "homolog" from Glycine max, 7 also form a cluster (cluster 2), suggesting that IFRH-Gm1 may in fact be an isoflavone reductase. Cluster 3, with the exception of IFRH-Cp1, from Citrus ϫ paradisi (48), which is an outlier, consists of two groups, corresponding to those present in the gymnosperms and those in the angiosperms. This, in turn, raises an interesting question about the origin of enan-tiospecificity among these reductases. Since PCBERs are apparently highly ubiquitous in the plant kingdom, they may have been the progenitor enzymes of both PLR and IFR (with the latter present thus far mainly in the Fabaceae (8)), having been conscripted into these latter biosynthetic pathways, with enantiospecificity evolving later.
Conclusion-The enzymes catalyzing phenylcoumaran benzylic ether reductions appear to be ubiquitous in the plant kingdom and were previously known as PLR and IFR "homologs," based on their high sequence homology to these two classes of related enzymes. The cloning and heterologous expression of the PCBER from P. taeda, together with that from P. trichocarpa (PCBER-Pop1), now identifies the first known catalytic function for this enzyme class. In addition, the regiospecific reduction of DDC 5 to IDDDC 6 gives the first biochemical proof that 8 -5Ј-linked lignans are the products of distinct biochemical pathways, unrelated to that involved in lignin biosynthesis. These data provide further evidence against the hypothesis that both phenolic coupling and subsequent metabolism in plants are random, uncontrolled processes, as was once imagined (52,53).
Having thus identified a catalytic function for these proteins, future studies will determine substrate versatility, the nature of the molecular species that is bound to the enzyme, and the mechanism of the subsequent transformation (i.e. hydride transfer to the quinone methide or some other sequence of reductive events). Additionally, future kinetic and inhibition studies are expected to establish proportions of active enzyme in the recombinant protein preparations. Interestingly, PCBERs effect regio-rather than enantiospecific conversions and may even be the progenitors of both PLRs and IFRs, which became, in contrast, enantiospecific. Moreover, with all three distinct classes of related reductases (PCBER, PLR, and IFR) available, an avenue of comparative mechanistic study is now available, i.e. in order to establish their points of similarity and difference, which permit some (PLR and IFR) to discriminate between substrate enantiomers, whereas others (PCBER) do not.