Digitalis purpurea P5βR2, encoding steroid 5β-reductase, is a novel defense-related gene involved in cardenolide biosynthesis

Summary

• •

  The stereospecific 5β-reduction of progesterone is a required step for cardiac glycoside biosynthesis in foxglove plants. Recently, we have isolated the gene P5βR, and here we investigate the function and regulation of P5βR2, a new progesterone 5β-reductase gene from Digitalis purpurea.

• •

  P5βR2 cDNA was isolated from a D. purpurea cDNA library and further characterized at the biochemical, structural and physiological levels.

• •

  Like P5βR, P5βR2 catalyzes the 5β-reduction of the Δ⁴ double bond of several steroids and is present in all plant organs. Under stress conditions or on treatment with chemical elicitors, P5βR expression does not vary, whereas P5βR2 is highly responsive. P5βR2 expression is regulated by ethylene and hydrogen peroxide.

• •

  The correlation between P5βR2 expression and cardenolide formation demonstrates the key role of this gene in cardenolide biosynthesis, and therefore in the chemical defense of foxglove plants.

Introduction

Although plant secondary metabolism is mostly related to defense mechanisms or interactions with the environment, interest in many plant secondary metabolites is mainly a result of their biotechnological applications. Plant sterol metabolism is of special interest, not only because of its involvement in the synthesis of compounds essential for plant growth and development, i.e. brassinosteroids, but also because plant sterols include economically important plant secondary products, such as cardenolides, phytoecdysteroids, glycoalkaloids, pregnane derivatives and saponins.

Cardenolides are a group of plant secondary metabolites, many of which possess potent pharmacological and other bioactive properties. At present, cardiac glycosides are widely used as inotropic drugs to improve hemodynamics in patients with compromised cardiac function. Foxglove species, mainly Digitalis lanata and D. purpurea, are cultivated around the world, as the chemical complexity of cardiac metabolites precludes their synthetic production.

Cardenolide biosynthesis requires the transformation of sterols to pregnenolone. Then, this pregnene is converted by Δ5-3β-hydroxysteroid dehydrogenase (3βHSD; EC 1.1.1.145) into isoprogesterone, which is isomerized to progesterone. Progesterone 5β-reductase (P5βR; EC 1.3.1.3) catalyzes the 5β-reduction of progesterone to 5β-pregnan-3,20-dione, and is considered to be the first committed step in the branch pathway leading to cardenolides (Fig. 1). In a second stereospecific reaction, 3β-hydroxysteroid-5β-oxidoreductase converts 5β-pregnan-3,20-dione into 5β-pregnan-3β-ol-20-one. Sequential hydroxylations at C14 and C21 and lactone ring formation at C17 lead to aglycone digitoxigenin. Stereospecific reactions determine that all the Digitalis cardenolides are β-configured at C3, C5, C14 and C17. The therapeutic action of cardiac glycosides depends not only on the structure of the distinct aglycones, but also on the type and number of sugar units attached at C3.

Figure 1

First steps of the putative biosynthetic pathway of cardenolides. (a) Schematic biosynthetic pathway. The specific cardenolide pathway is marked in bold, showing enzymes for which the corresponding genes have been isolated. (b) First step in the specific cardenolide biosynthetic pathway: 5β-reduction of progesterone by P5βR isoenzymes.

P5βR is considered to be a key enzyme in cardenolide biosynthesis as: it is the first stereospecific enzyme of the pathway leading to 5β-configured derivatives; it appears to be the initial step in cardenolide biosynthesis (cholesterol and pregnenolone feeding do not stimulate cardenolide accumulation, whereas progesterone and 5β-pregnan-3,20-dione lead to an increase by 180 and 200–300%, respectively; Gärtner & Seitz, 1993); and it is the only known enzyme which is exclusively present in shoot cultures and young plants (Gärtner & Seitz, 1993) (cardenolide accumulation coincides with shoot and green plantlet formation; Diettrich et al., 1990).

Cloning the gene that encodes this key enzyme was successfully undertaken by purification of the native P5βR from D. purpurea leaves and subsequent partial amino acid sequencing (Gärtner et al., 1994; Gavidia et al., 2007). Attempts to clone P5βR cDNA using a homology-based strategy with a Δ⁴-3-ketosteroid 5β-reductase from rat failed (Gavidia et al., 2002). P5βR is ubiquitous in plants and belongs to the short-chain dehydrogenase/reductase (SDR) superfamily. As P5βR differs from known SDRs at catalytically essential sites, and novel active-site motifs are highly conserved across plant and bacterial P5βR homologs, we proposed that P5βR operates via a novel reaction mechanism defining a novel subfamily of extended SDRs, or even a new SDR family (see Gavidia et al., 2007). Unlike plants, the 5β-reduction of progesterone in vertebrates is accomplished by a member of the aldo-keto reductase superfamily; therefore, P5βR illustrates a good case of a metabolic function for which animals and plants have independently recruited unrelated genes (Gavidia et al., 2007).

The occurrence of multiple genes encoding a specific enzyme is a general feature of higher plants. Previous results from our laboratory have suggested that P5βR is a member of a small multigene family of two presumptive P5βR copies and of at least two other copies with related sequences (Gavidia et al., 2007). In this work, we report the isolation, biochemical characterization and expression pattern of the gene P5βR2, a D. purpurea P5βR isoenzyme. A structural model of P5βR2 and a feasible reaction mechanism of this protein are depicted. We have also carried out a comparative study into the regulation of P5βR and P5βR2 gene expression and their role in cardenolide biosynthesis.

Materials and Methods

Plant material

Foxglove (Digitalis purpurea L.) seeds were germinated, and plants were grown in a glasshouse in pots containing a mixture of soil and vermiculite (3 : 1) under a 16-h light : 8-h dark cycle at 22°C. Plants were watered daily and supplemented once a week with nutrient solution. Unless otherwise stated, the different samples used in the experiments described above were harvested, washed, frozen in liquid nitrogen and stored at −80°C until use.

Cloning of P5βR2 cDNA

A D. purpurea cDNA library was screened under moderate stringency conditions. Nylon filters were prehybridized and hybridized at 50°C in 330 mM sodium phosphate buffer (pH 7), 7% sodium dodecylsulfate (SDS), 1 mM ethylenediaminetetraacetic acid (EDTA) and 1% bovine serum albumin, and, after overnight incubation, were washed twice for 15 min in 2 × standard saline citrate (SSC) with 0.1% SDS at 50 and 40°C, consecutively. The use of a genomic fragment (850 bp) corresponding to P5βR as a probe (see Gavidia et al., 2007) allowed the isolation of several positive clones other than P5βR. Their cDNA inserts were in vivo excised and subcloned into pBluescript SK(−) phagemid vector, and then sequenced using an ABI 310 sequencer (Perkin-Elmer, Wiesbaden, Germany). For the isolation of the entire P5βR2 cDNA, PCR was performed using the D. purpurea cDNA library as a template, a universal 5′-primer (T3) of the library and a specific 3′-primer (5′-CTTCCCTTCATAACCCAC-3′) based on the sequence of the cDNA of interest truncated at the 5′ terminus. The screening of a D. purpurea genomic library (see Gavidia et al., 2007), using the entire P5βR2 cDNA as a probe, led to the isolation of the P5βR2 gene.

Heterologous expression

The full-length open-reading frame of P5βR2 cDNA was subcloned into the XmnI-PstI sites of pMAL-c2 (New England Biolabs, Frankfurt, Germany). It was expressed as a fusion protein with maltose-binding protein at the N-terminus. The PstI-ended fragment was amplified using the primers 5′-ATGTATACCGACACAACGACTTGG-3′ and 5′-TTTTTCTGCAGTCAAGGGACAAATCTATAAG-3′, and Pfu DNA polymerase (NEED S.L., Valencia, Spain). The constructed vector was transformed into Escherichia coli BL21(DE3)pLysS and cultured in Luria–Bertani medium supplemented with glucose (2 g l⁻¹) and ampicillin (100 mg l⁻¹). Gene expression was induced by the addition of 0.3 mM isopropyl-β-d-thiogalactopyranoside at very early log phase culture [optical density at 600 nm (OD₆₀₀) = 0.1], and then incubated at 4°C for 3 d. Cells were harvested by centrifugation, disrupted by sonication, and the soluble maltose-binding protein-tagged P5βR2 protein was then purified to apparent homogeneity as described previously by Gavidia et al. (2007).

Enzyme assays and GC–MS analysis

After purifying the recombinant protein, it was concentrated and the elution buffer was changed for a reaction buffer containing 50 mM Tris–HCl, pH 7.5, 250 mM sucrose, 2 mM EDTA and 1 mM dithiothreitol, using Vivaspin concentrators (VIVASCIENCE, Stonehouse, UK). The protein was quantified by the method described by Bradford (1976). In the P5βR2 activity assays, and in accordance with Gärtner et al. (1994), the standard reaction mixture consisted of 6.4 mM NADP+, 32.1 mM glucose 6-phosphate, 42 nkat glucose 6-phosphate dehydrogenase, 270 μM progesterone and 100 μg of recombinant P5βR2 protein, in a final volume of 750 μl in the reaction buffer. Assays were incubated for 3 h at 30°C, and the reaction was terminated by the addition of CH₂Cl₂. The extraction and purification of the pregnanes were carried out as described by Gärtner & Seitz (1993). Pregnanes were analyzed by GC–MS as described by Gavidia et al. (2007). All the steroid substrates employed were purchased from Sigma with the highest purity available.

External treatments of D. purpurea plants

To determine the effects of several stress conditions and chemicals on the accumulation of the P5βR and P5βR2 transcripts, foxglove plants were grown in a glasshouse over a 2-month period.

For the mechanical wounding experiments, 1-mm-diameter holes were made across the lamina, which effectively damaged c. 5% of the leaf area. Samples were collected 1, 2, 3, 6 and 15 h following treatment. In a second experiment, samples were also collected at longer time periods after wounding (1, 2, 3, 6, 8, 15, 24, 48 and 72 h). Low-temperature treatment was carried out by exposing the plants grown at 22°C to a temperature of 4°C, and leaf samples were harvested after 2 and 4 d. The plants subjected to salt stress were watered with a solution of 250 mM NaCl, and samples were collected after 6, 24 and 48 h. Leaves were also sampled from heat-shocked plants which were incubated at 41°C for 1, 2 and 4 h.

For the chemical treatments, plants were sprayed in separate cabins with the following solutions (plus 0.1% Tween-20): 100 μM methyl jasmonate (MJ) in 0.1% ethanol, 2 mM salicylic acid (SA), 0.5 mM 1-aminocyclopropane-1-carboxylic acid (ACC) and 20 mM hydrogen peroxide (H₂O₂). Control plants were sprayed with 0.1% Tween-20 in water or 0.1% ethanol. Leaves were collected at 4, 8, 24 and 48 h after treatment.

In all the treatments, the collected leaves were immediately frozen in liquid nitrogen and stored at −80°C until used for RNA extraction. All the chemicals employed were obtained from Sigma with the highest purity available.

Transcript analysis

RNA extraction from several tissues was performed according to the method described by Steimle et al. (1994), except for root samples where this method was modified as described by Roca-Pérez et al. (2004). Before reverse transcription, total RNAs were treated with DNase I (Takara, Otsu, Shiga, Japan), according to the manufacturer’s protocol, to remove any potential DNA contamination. Five micrograms of total RNA were used for cDNA synthesis. Reverse transcription was performed using 50 units of SuperScript II RT (Invitrogen, Carlsbad, CA, USA) in the presence of 40 units of recombinant ribonuclease inhibitor (Invitrogen). After first-strand synthesis, the RNA template was removed by digestion with 2 units of RNase H (Invitrogen). For semi-quantitative reverse transcriptase-polymerase chain reaction (RT-PCR), cDNA derived from 0.5 μg RNA was used for PCR with P5βR- or P5βR2-specific primers and the following programme: 94°C for 5 min, 30 cycles at 94°C for 30 s, 55°C for 30 s and 72°C for 1 min, and final incubation at 72°C for 10 min. PCR products were analyzed on 1% agarose gel. [Note: the specific primers used in both RT-PCR and real-time quantitative PCR (RT-qPCR) analyses will be provided on request.]

A quantitative analysis of transcript expression was carried out by RT-qPCR using the SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA, USA) following the standard manufacturer’s protocol. After RNA extraction and reverse transcription, 300 ng of first-strand cDNA were used for real-time PCR with the P5βR-, P5βR2- or 3βHSD-specific primer. The results were normalized to the reference gene actin using StepOne software v2.0 (Applied Biosystems), and were always expressed as expression ratios relative to a control value.

Cardenolide extraction and determination

Leaves were surface cleaned with distilled water, dried at 50°C for 48–72 h and pulverized. Cardenolides were extracted as described by Roca-Pérez et al. (2004) and analyzed by high-performance liquid chromatography according to Gavidia & Pérez-Bermúdez (1997). Extracts were analyzed in a Merck-Hitachi LaChrom chromatograph (Tokyo, Japan) (L-7400 UV detector, L-7100 pumps, L-7200 auto-sampler) coupled to a 20-μl injector (Rheodyne 7725). Cardenolides were separated at 25°C on a LiChroCART 250-4 column packed with LiChrospher 100 RP-18 (5 μm) from Merck, and detected at 230 nm under a flow rate of 1 ml min⁻¹. A gradient elution of H₂O (A) and CH₃CN (B) was employed: initial, 20% B; 35 min, 32% B; 45 min, 40% B; 55 min, 50% B; 59 min, 55% B; 65 min, 60% B; 70 min, 20% B. The identities and amounts of cardenolides were checked by co-chromatography with commercially available standards (Sigma). The standards employed were from series A [digitoxigenin, digitoxigenin-monodigitoxoside (evatromonoside), digitoxigenin-bisdigitoxoside, digitoxigenin-tridigitoxoside (digitoxin) and lanatoside A] and B [gitoxigenin, gitoxigenin-tridigitoxoside (gitoxin) and lanatoside B]. We also included α- and β-acetyldigitoxin as a reference of possible alterations during plant material processing. Four replicates were analyzed for each experimental sample. The final data are presented as the number of milligrams of cardenolides per gram of dry weight.

Modeling of the P5βR2 structure

The structural model for P5βR2 was calculated using the Discovery Studio Modeling software package (Accelrys, San Diego, CA, USA). The structure of progesterone 5β-reductase from Digitalis lanata in a complex with NADP (Protein Data Bank code 2V6G) was identified as the best possible template from the Protein Data Bank. Both the P5βR2 32-394 fragment and 2V6G exhibit a sequence identity of 55.9% and a sequence similarity of 72.5%, calculated using a BLOSUM substitution matrix (Henikoff & Henikoff, 1992). Initial structural models were calculated by aligning the P5βR2 sequence with the template by the MODELLER Align2D method, followed by the satisfaction of spatial constraints (Sali & Blundell, 1993). The parts of the model sequence which were not aligned to the template structures were optimized with the statistical pair potential DOPE (Shen & Sali, 2006) within MODELLER. Consequently, 20 structural models were built for P5βR2 from which that with the best DOPE score was selected for further evaluation. Structural verification of the selected model was determined by the Verify Protein protocol within Discovery Studio, Verify3D (Lüthy et al., 1992), which measures the compatibility of an amino acid sequence with a three-dimensional protein structure and PROCHECK (Laskowski et al., 1993).

NADP coordinates were obtained from progesterone 5β-reductase NADP complex and progesterone coordinates from the human progesterone receptor complex (Protein Data Bank code 1A28). An initial P5βR2 NADP complex was obtained by the superposition of P5βR2 and 2V6G structures and manual fitting. From this initial NADP pose in P5βR2, a binding site for the cofactor, based on its volume, was defined, and NADP docking was explored using LigandFit (Venkatachalam et al., 2003). Progesterone docking was explored in a similar way, starting with a pose placed in the cavity next to the cofactor and by taking into account the Δ⁴ double bond location with regard to the nicotinamide ring of NADP, as described by Thorn et al. (2008). The Piecewise Linear Potential 1 function (Gehlhaar et al., 1995) and the standard parameters in Discovery Studio were used in all the docking calculations.

Results

Cloning and sequence analysis of P5βR2

The screening of a cDNA library generated from D. purpurea leaves (Gavidia et al., 2002), under moderate stringency conditions, led to the isolation of different positive clones. One of the cDNA clones contained a fragment of c. 900 bp. The corresponding full-length cDNA was isolated from the cDNA library by PCR using a 5′-nonspecific primer (T3 primer) and a sequence-specific antisense primer complementary to the cDNA fragment. The resulting clone, designated P5βR2, has a length of 1332 bp and contains a 1182-bp-long open reading frame. The screening of a D. purpurea genomic library also led to the isolation of the P5βR2 gene. The analysis of the sequence of this genomic clone demonstrated P5βR2 to be a full-length cDNA clone, and revealed the presence of a unique intron (86 bp long) located 57 bp from the ATG codon. Nucleotide sequence data is available in the GenBank database under accession number GU062787.

The P5βR2 cDNA was deduced to encode a polypeptide of 394 amino acids with a calculated molecular weight of 44.2 kDa, similar to that of P5βR (M_r = 43 963 Da), and a theoretical isoelectric point of 5.11. The nucleotide sequences of P5βR and P5βR2 from D. purpurea exhibited 63% identity, and their amino acid sequences presented 54% identity (Fig. 2).

Figure 2

Alignment of the deduced amino acid sequence of P5βR (AJ310673) and P5βR2 (GU062787, this study). The locations of motifs 1–8 are shown with bars.

A database search revealed no significant homology to other proteins other than those corresponding to P5βR. Unlike the ubiquity of P5βR, which is highly conserved throughout the plant kingdom (Gavidia et al., 2007), the presence of P5βR2 is restricted to certain species, including those of the genera Digitalis and Isoplexis (Tarrío R, Rodríguez-Trelles F, Pérez-Bermúdez P & Gavidia I, unpublished results).

Heterologous expression of P5βR2

To examine the catalytic function of P5βR2, the corresponding cDNA was expressed in E. coli as a fusion with maltose-binding protein, and was purified by affinity chromatography using amylose resin. Heterologous protein induction was carried out at different temperatures (37 to 4°C) and cell densities (OD₆₀₀ of 0.6 to 0.1). Among all the conditions tested, an induction at very early log phase cultures under low temperatures (OD₆₀₀ = 0.1 and 4°C) allowed a high rate of production of the heterologous P5βR2. Moreover, the expression of endogenous bacterial proteins was minimized, thus making the protein purification process more amenable and efficient. The recombinant protein was soluble and catalytically active with an estimated molecular mass of 87 kDa (Fig. 3).

Figure 3

Heterologous expression of P5βR2 cDNA in Escherichia coli. SDS-PAGE analysis of total protein extracts (15 μg) from induced E. coli cells at early (0.1) or mid (0.6) log phase culture (optical density at 600 nm, OD₆₀₀). The cells were grown at 4°C for 1, 2 or 3 d after induction with 0.3 mM isopropyl-β-d-thiogalactopyranoside. Purified, recombinant P5βR2 is also shown. M, markers.

The purified P5βR2 showed significant steroid 5β-reductase activity by catalyzing the reduction of progesterone to produce 5β-pregnane-3,20-dione, as demonstrated by GC and MS analyses. Like P5βR, the recombinant form of P5βR2 can catalyze the reduction of several steroids with a 3-oxo,Δ^4,5 structure, although the highest substrate specificity was obtained with progesterone. The steroid 5β-reductase activity of P5βR2 requires NADPH, rather than NADH, as a cofactor, and the catalyzed reaction was irreversible. To analyze the biochemical properties of the expressed P5βR2 protein, 20 μg of recombinant protein were incubated at 30°C with varying amounts of NADPH (0–150 μM) or progesterone (0–300 μM). The apparent K_m value for progesterone was determined to be 29.9 μM with V_max = 29.5 μkat kg⁻¹ (Fig. 4a), whereas the apparent K_m value for NADPH was 21.4 μM with V_max = 39.4 μkat kg⁻¹ (Fig. 4b). Optimal progesterone reduction by the recombinant enzyme was observed at pH 7.0–7.5. When comparing the kinetic parameters of P5βR2 with those of P5βR (Gavidia et al., 2007), the V_max/K_m ratios indicated that P5βR exhibits higher specificity towards NADPH and progesterone than does P5βR2, although P5βR2 presents a greater apparent affinity to progesterone than does P5βR (Table 1).

Figure 4

Biochemical characterization of recombinant P5βR2. The purified enzyme (20 μg) was incubated with varying concentrations of progesterone (0–300 μM) or NADPH (0–150 μM) to determine their respective kinetic data. Activity was calculated on the basis of the reduction in absorbance at 340 nm. Hanes plots for the substrate (a) and cofactor (b).

Table 1. Compared kinetic parameters of recombinant P5βR and P5βR2

	Progesterone			NADPH
	V max (μkat kg^−1)	K m (μM)	V max/Km	V max (μkat kg^−1)	K m (μM)	V max/Km
P5βR^1	57	39.1	1.5	94.7	7.9	12.0
P5βR2	29.5	29.9	1.0	39.4	21.4	1.8

• ¹Values were taken from Gavidia et al. (2007).

P5βR2 structure and reaction mechanism

A primary structural analysis of the P5βR2 protein revealed the presence of the motifs that are characteristic of the new family of SDRs represented by P5βR as a prototype (Gavidia et al., 2007). An example is NFYYxxED (motif 6), which we considered to be a proxy of the typical SDR catalytic motif YxxxK (Fig. 2; Gavidia et al., 2007), indicating that P5βR2 belongs to the same SDR family as P5βR. In addition to the specific motifs of this new SDR family, it exhibits conserved sequence motifs that resemble patterns of the ‘extended’ and ‘classical’ SDR families: GxxGxxG (motif 1) and Dhx[cp] (motif 3), respectively (Fig. 2; see Persson et al., 2003).

A structural model for P5βR2 of D. purpurea (Fig. 5a) was obtained by homology using the X-ray structure of D. lanata 5β-POR bound to the cofactor as a template. The structure obtained shows typical structural features, such as the NADPH binding site and the Rossman fold. With this structural model, we have been able to attribute hypothetical roles to some new sequence motifs that are not found in other SDRs (see Table 2).

Figure 5

Three-dimensional structural model and reaction mechanism of P5βR2 from Digitalis purpurea. (a) Ribbon representation of the enzyme showing α-helices in blue and β-sheets in purple; NADP cofactor (orange carbons) and progesterone substrate (green carbons) are shown as stick representations. (b) Proposed reaction mechanism for the reduction of the Δ⁴ double bond of progesterone in P5βR2.

Table 2. Sequence motifs in the restricted short-chain dehydrogenase/reductase (SDR) family

Motif	Location	Function
GxxGxxG	Loop β1 to α1	Coenzyme binding (recognition of pyrophosphate and ribose linked to the adenine ring)
GxxRR	Final of β2 and loop to β3	O2′-phosphate binding, discriminates between NADH and NADPH
Dhx[cp]	Loop β2 to α2	Interaction with the adenine ring of NADPH
DhTHhFYVpW[hp]	β4 and loops to α2 and α3	Contains central β-strand. Structural role in stabilizing central β-sheet
TGxKxYhG[hp]	Loop β5 to α5	Contains catalytic lysine residue
NFYYxxED	α5	Contains catalytic tyrosine residue
Wp[VI]HRP	β6	Part of β-sheet. Proline packs against nicotinamide ring of NADPH
D[hp]DLhA[ED]	α8	Electrostatic driving force for the reaction

• In the motifs, ‘c’ denotes a charged residue, ‘h’ a hydrophobic residue, ‘p’ a polar residue and ‘x’ any residue. Alternative amino acids at a motif position are shown in brackets.

Interestingly, there is a new sequence motif involved in the Rossman double fold. The DhTHhFYVpW[hp] motif is found in the central β-strand of the β-sheet (β4), and is also part of the loops connecting this strand to the previous and posterior α-helices (α2 and α3) (Fig. 2, motif 4). Some of the residues contained in this motif make hydrophobic contacts with residues of the surrounding α-helices. Thus, this motif is expected to contribute to the structure of not only the β-sheet, but of the whole enzyme. Another new motif is D[hp]DLhA[ED] (Fig. 2, motif 8) found in α8. This motif, which is rich in negatively charged residues, could play a role in assisting the hydride transfer from the cofactor to the substrate.

In our model, hydrophobic interactions between the enzyme and the substrate are established through residues I138, W152, M155, L210, V253, F348, V351 and A352. In the case of progesterone, the C4–C5 (Δ⁴) double bond of the substrate is conjugated with the C3 carbonyl group. Thus, the reduction of this double bond can take place through a 1,4-addition mechanism. A hydride is transferred from NADPH to the C5 atom and, subsequently, the substrate is protonated at the oxygen position of the C3 carbonyl group, resulting in the corresponding enol intermediate (Filling et al., 2002). A fast keto–enol tautomerization process could then yield the final product of the reaction. The characteristic motif for this new SDR family, NFYYxxED, seems to be involved in proton transfer to the C3 carbonyl group. This motif contains the important catalytic Y183 and is placed in α5. This residue is equivalent to Y179 found in P5βR; mutations of this residue to alanine and phenylalanine resulted in nondetectable enzymatic activity (Thorn et al., 2008). Our structural model suggests that Y183 is a suitable candidate to protonate directly the oxygen atom of the progesterone carbonyl group, whereas, in the model of Thorn et al. (2008), the cofactor was proposed as the proton donor. The acidity of this tyrosine could be enhanced by a hydrogen bond with the hydroxyl group of the ribose of the cofactor that is immediately placed after the nicotinamide ring. More importantly, a positively charged lysine (K149) appears to be close enough in the active site and could stabilize the basic form of Y183, thus lowering its pKa value. It is precisely this lysine residue that forms part of another new sequence motif, TGxKxYhG[hp], located in the loop connecting β5 to α5 (Fig. 2, motif 5). In addition to the role it plays in enhancing the acidity of Y183, this residue also acts by favoring hydride transfer from NADPH to the substrate. Effectively, hydride transfer involves the migration of a negative charge from the cofactor to the substrate; thus, the positive charge of K149, which is closer to the latter than to the former, must provide an important driving force for the reaction. Figure 5b shows a schematic view of the reaction mechanism and the role played by the aforementioned residues.

Accumulation of P5βR and P5βR2 transcripts

The expression pattern of gene P5βR2 in several D. purpurea tissues was analyzed by semi-quantitative RT-PCR and the results were compared with those of P5βR (Gavidia et al., 2007). The P5βR2 and P5βR transcripts were discriminatively amplified using specific primers for the respective cDNAs, and specific primers for actin were also used as controls for the constitutive expression. RNA was extracted from mature flowers, leaves, stem at the base of the rosette, vascular bundle at the basal part of the leaf and roots. As in our previous analysis for P5βR, the RNA transcripts of P5βR2 were present in all the organs tested, and were seen to be more abundant in leaves and flowers. Unlike P5βR, which presented the lowest amount of transcripts in the vascular bundle, the mRNA of P5βR2 in roots was only detected when the number of PCR temperature cycles was increased (Fig. 6a). To compare the expression of both genes, we determined their relative transcript levels in relation to the expression of the reference gene actin by RT-qPCR using the cDNA from leaves as a template. These data are presented as a percentage in relation to the expression levels found for P5βR, as its expression is 16-fold higher than that of the reference gene. Figure 6b illustrates that P5βR presents the higher expression of the two genes, where P5βR2 presents 1.5% of the P5βR expression.

Figure 6

Comparative expression of the P5βR and P5βR2 genes. (a) Semi-quantitative reverse transcriptase-polymerase chain reaction (RT-PCR) analysis in flowers (F), leaves (L), stem (S), vascular bundles (V) and roots (R). RT-PCR of the actin gene was used as a loading control. (b) Relative steady-state transcript levels of P5βR and P5βR2 in leaves of Digitalis purpurea measured by real-time quantitative PCR (RT-qPCR).

Differential and inducible expression of P5βR and P5βR2

Secondary metabolites play a key role in the interaction between plants and their environment (Kutchan, 2001). For example, the pigments produced in flowers to be pollinated attract insects. Other metabolites are involved in plant defense; therefore, their biosynthesis is frequently affected by abiotic and biotic stresses. It is well established that cardenolides play a role in plant defense by acting as deterrents to herbivores; therefore, the presence of two P5βR genes led us to study their transcription profiles under different stress conditions. We used 2-month-old D. purpurea plants grown in soil to examine the changes in the steady-state levels of P5βR and P5βR2 RNAs following mechanical leaf injury (wounding), high and low temperatures, and NaCl treatment (salt stress). As shown above, intact Digitalis leaves contained much lower levels of P5βR2 than P5βR RNA (6, 7). When plants were subjected to the aforementioned treatments, the level of P5βR transcripts revealed by semi-quantitative RT-PCR analysis remained constant over all the time points (Fig. 7), whereas the expression of P5βR2 increased in relation to untreated plants. When plants were maintained at low temperatures (4°C), the maximal level of expression was obtained after 2 h, with the signal decreasing by 4 h. At high temperatures (41°C), a lower but detectable increment in the level of expression could be observed at 1 h and 2 h, which decreased until the control levels after 4 h. The upregulation of P5βR2 expression was clearly apparent after 24 h under salt stress, reaching a maximum at 48 h (Fig. 7).

Figure 7

Semi-quantitative reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of P5βR and P5βR2 expression in response to wounding, heat shock, cold shock and salt stress in Digitalis purpurea leaves. Actin expression was analyzed as a control. C, untreated plant.

7, 8 show the changes in the steady-state levels of P5βR2 RNA following leaf injury, where the data presented correspond to the damaged leaf. A time course until 15 h was analyzed by semi-quantitative RT-PCR (Fig. 7). It clearly showed two different responses. The first was a peak of transcript accumulation with maximum levels at 1 h, followed by a rapid decrease, as, at 2 h, the signal was only slightly higher than the control. The second response was slower, but was maintained throughout the time points, as the number of P5βR2 transcripts increased gradually from 2 h until 15 h. Given the role of cardenolides as a deterrent for Digitalis plants of nonadapted herbivores, we carried out a second experiment by taking samples at the same and longer time points to emphasize the responses of P5βR and P5βR2 to wounding. In this experiment, we determined the relative transcript level to the reference gene expression by RT-qPCR over longer time periods. As Fig. 8 illustrates, the upregulation of P5βR2 expression remained evident after 24 h of injury, whereas, at 48 h, the increment in relation to the control was not significant. Among all the time points tested, the P5βR2 RNA levels reached their maximum at c. 1 h and 15 h after wounding. P5βR transcripts were also unaffected by wounding over these longer periods, until 72 h after injury.

Figure 8

Expression analysis of P5βR (black bars), P5βR2 (grey bars) and 3βHSD (open bars) in response to wounding. Real-time quantitative polymerase chain reaction (RT-qPCR) analysis of wounded plants of Digitalis purpurea at 1, 2, 3, 6, 8, 15, 24, 48 and 72 h. Results are normalized to the actin transcript and are presented as the relative expression compared with control. C, unwounded plant. The average values of three independent determinations of the transcription levels are presented with error bars indicating ± SE.

The differential expression of P5βR and P5βR2 in response to environmental stresses led us to determine whether the signaling molecules SA, MJ, H₂O₂ and ethylene induced the expression of P5βR, and which were involved in the induction of P5βR2. Figure 9 shows that the expression of P5βR above the control (untreated plants) was not observed by the application of exogenous SA, MJ, H₂O₂ or the precursor of ethylene: ACC. By contrast, P5βR2 expression was induced by H₂O₂ and ACC, but was independent of SA and MJ.

Figure 9

Expression analysis of P5βR (open bars) and P5βR2 (grey bars) in response to elicitors. Transcripts were monitored by quantitative real-time reverse transcriptase-polymerase chain reaction (RT-PCR) analysis using Digitalis purpurea plants treated with salicylic acid (SA), 1-aminocyclopropane-1-carboxylic acid (ACC), methyl jasmonate (MJ) and hydrogen peroxide (H₂O₂). The results are normalized to the actin transcript and are presented as the relative expression compared with the controls. C, untreated plants. The average values of three independent determinations of the transcription levels are presented with error bars indicating ± SE. Insets for SA and MJ are also depicted at a reduced scale to clarify the lack of P5βR and P5βR2 induction by these signaling molecules.

Cardenolide biosynthesis under wounding

The variation of cardenolide content under different environmental stresses and chemical elicitors has been studied in Digitalis and Asclepias species in the last few decades. It has been shown that several hormones (Ohlsson, 1990; Berglund & Ohlsson, 1992; Palazón et al., 1995), CO₂ and water stress (Stuhlfauth et al., 1987), salinity (Morales et al., 1993) and H₂O₂ (Paranhos et al., 1999) alter the synthesis of cardenolides, and usually increase their production. As cardenolides act as a direct plant defense against herbivores, the production of these secondary metabolites has been examined in Asclepias after feeding by adapted specialists (monarch butterfly and weevil) and wounding (Malcolm & Zalucki, 1996; Fordyce & Malcolm, 2000).

We qualitatively and quantitatively analyzed the content of cardenolides after wounding in 2-month-old D. purpurea plants. Two plants were wounded and cardenolides were extracted twice from each plant for each time point. Given the large variation in cardenolide content usually found between individual Digitalis plants (Gavidia et al., 1996), we excised some leaves at the initial time point (t = 0), which were used as a control for each plant. The remaining leaves were immediately wounded. The results were expressed as the variation in cardenolide in relation to undamaged control leaves. Figure 10 shows the variation in all the cardenolides analyzed (total cardenolides hereafter) and two specific cardenolides of series A, digitoxigenin tridigitoxoside (digitoxin) and digitoxigenin monodigitoxoside (evatromonoside, DM hereafter), which were the most abundant in D. purpurea extracts. Wounding favored total cardenolide biosynthesis with a maximum increase after 4 h, which approximately decreased to the control levels at 6 h. From 6 h to day 5 cardenolides increased, but the increment was slower and lower than that observed at 4 h. Such changes in the total cardenolide content were mainly caused by the increase in DM content (Fig. 10). Indeed, the levels of digitoxin after wounding diminished in comparison with the control and did not reach the control values until 7 d after treatment (Fig. 10). These data suggest a dual effect of wounding on cardenolide biosynthesis: an increase in the less polar cardenolides (DM) in the short and long term, which correlates with the transcription profile of P5βR2 (see also Fig. 8), and a reduction in the content of the most polar cardenolides (digitoxin).

Figure 10

Analysis of cardenolides in response to wounding. Total cardenolides (diamonds), digitoxigenin monodigitoxoside (DM; squares) and digitoxin (triangles) were determined in wounded leaves of Digitalis purpurea plants. Each value represents the cardenolide variation in relation to its own control, unwounded leaves at time 0, as described in Materials and Methods. The values are the means ± SE (n = 3).

Other genes putatively involved in cardenolide biosynthesis

3βHSD, lanatoside 15′-O-acetylesterase (LAE) and cardenolide 16′-O-glucohydrolase (CGH) are, in addition to the P5βRs, the only known genes from Digitalis that are putatively involved in different cardiac glycoside metabolism steps. LAE and CGH do not participate in the biosynthesis of the genin, but in cardiac glycoside catabolism, as their activities lead to modifications of the glycosyl chain attached to the genin skeleton. Such modifications are supposed to be determinant in the intracellular and long-distance transport of cardenolides. CGH may also play an important part in the defense against predators (Framm et al., 2000).

It has been demonstrated that the enzyme 3βHSD converts pregnenolone into isoprogesterone. However, certain controversy still exists as to whether this enzyme is responsible for progesterone formation through isoprogesterone isomerization (Finsterbusch et al., 1999; Lindemann et al., 2000; Herl et al., 2007). We monitored the expression level of a 3βHSD gene (GenBank: AY789453) from D. purpurea following mechanical leaf injury in 2-month-old plants. As Fig. 8 shows, the results reveal that the level of 3βHSD transcripts remained almost invariable at all the time points analyzed when compared with the levels obtained for P5βR2. Therefore, among the genes known to be involved in cardenolide biosynthesis, only the expression pattern of P5βR2 is correlated with the variation in cardenolide contents of wounded plants.

Discussion

The aim of this study was the isolation and characterization of new cDNAs which might be implicated in cardenolide biosynthesis in Digitalis species. In our previous work, two approaches were used for the isolation of P5βR cDNA. First, a homology-based strategy (using rat Δ⁴-3-ketosteroid 5β-reductase), which failed (Gavidia et al., 2002) as P5βR was evolutionarily unrelated to its mammalian counterpart (Gavidia et al., 2007). Second, the purification of the native protein from D. purpurea leaves followed by amino acid sequencing and cDNA cloning (Roca-Pérez et al., 2004; Gavidia et al., 2007). As a result, we isolated a cDNA that encoded a protein responsible for P5βR activity. A comparative sequence analysis (Gavidia et al., 2007) showed that P5βR belongs to the SDR superfamily. Nevertheless, P5βR differs from known SDRs at catalytically essential sites.

An analysis of the genomic complexity of this novel P5βR gene family suggested that it comprises a small multigene family (Gavidia et al., 2007). In this work, we carried out a new screening of the D. purpurea cDNA library at moderate stringency to isolate cDNAs coding for members of this family. One of the isolated cDNAs, whose encoded protein showed the lowest similarity to P5βR (54%), catalyzes the same reaction, that is, the reduction of progesterone to 5β-pregnane-3,20-dione. This new member of the P5βR family, termed P5βR2, contains several conserved sequences of the SDR superfamily, as well as the novel motifs specific for the new family. Sequence analysis of the new P5βR family members introduced changes into the motifs determined earlier, thus demonstrating that this is indeed a new family, as predicted in Gavidia et al. (2007).

Persson et al. (2009) have established a functional subdivision of the SDR superfamily into at least 200 SDR families: the P5βR family has been designated as SDR75U, where U corresponds to the SDR type ‘Unknown’. Each of the SDR types is characterized by type-specific sequence patterns at the coenzyme binding site and/or the active site (Persson et al., 2009). The P5βR family belongs to a different type of SDR that is clearly distinguished by sequence patterns at the active site region. The new members of this family are restricted to some plant lineages and we propose the name ‘Restricted SDR type’. Following the nomenclature of Persson et al. (2009), the family designation should be SDR75R and the protein designations SDR75R1 and SDR75R2 for P5βR and P5βR2, respectively.

P5βR2 is a defense-related gene

A BLAST search in the database revealed no similarities of P5βR2 with proteins from any organism other than those corresponding to P5βR. However, we cloned P5βR2 homologs from different Digitalis and Isoplexis species (Tarrío R, Rodríguez-Trelles F, Pérez-Bermúdez P & Gavidia I, unpublished results). Therefore, it is plausible that this gene is restricted to certain plant lineages. The current distribution of P5βR2 in cardenolide-producing species differs from the ubiquity of P5βR, which is also widespread in noncardenolide-producing species (Gavidia et al., 2007). This strongly suggests that P5βR has physiological functions other than those derived from its involvement in cardenolide biosynthesis and, on this basis, P5βR2 could be closely associated with the formation of cardenolides.

In secondary metabolism, it is likely that multigene families encode isoenzymes with very similar biochemical properties, and that the essential distinction between them relates to their induction, or to their tissue or organ distribution (Walton et al., 1999). For P5βR/ P5βR2, both isoenzymes presented similar biochemical properties and substrate specificity measured in vitro. Both transcripts were present in all the tissues and organs tested, although the low steady-state level of P5βR2 mRNA, especially in roots, was noteworthy in relation to P5βR (Fig. 6). Nevertheless, these genes were expressed distinctly in relation to environmental cues, including wounding in particular. Under stress conditions, P5βR did not exhibit variation in its expression level, whereas the P5βR2 gene was responsive to cold, high salt, heat shock and wounding. Signaling molecules, such as SA, MJ, H₂O₂ and ethylene, accumulate during defense and, in turn, trigger the coordinated expression of defense genes. By comparing the expression of the P5βR and P5βR2 genes on treatment with such elicitors, it is clear that they are differentially regulated, as P5βR remains at constitutive levels, whereas P5βR2 expression is regulated by ethylene and H₂O₂. The pattern of expression and induction of P5βR2 resembles that of stress genes, which are normally silent and rapidly induced by stress conditions (see Cheong et al., 2002). Transcriptional profiling in wounded Arabidopsis has demonstrated the interaction between wounding and abiotic stress factors, as several transcription factors that govern stress regulation are among the early responsive genes after wounding (Cheong et al., 2002). Such interactions could explain why environmental factors activate a gene involved in the plant chemical defense against herbivores.

P5βR2 is a key gene for cardenolide biosynthesis induction

Cardenolides represent an interesting case of integral interaction between plants and their environment, especially in the field of chemical ecology. For example, the monarch butterfly is an adapted specialist that sequesters toxic cardenolides as larvae-feeding Asclepias sp. (cardenolide-producing species), and the adult is then chemically defended against predators. In response to insect damage, A. syriaca plants increase leaf cardenolide levels, which peak at 24 h after wounding, and are avoided by large herbivores (see Malcolm & Zalucki, 1996).

In Digitalis, we detected a maximum level of cardenolide content at 4 h after wounding, which decreased to constitutive levels after 6 h, and was then followed by a gradual and smaller increase in cardenolides until day 5 after wounding. We found that the pattern of total cardenolide biosynthesis on wounding correlates with the pattern of induction of P5βR2 under the same conditions. That is, a rapid and high initial induction that decreases rapidly to constitutive levels, followed by a slow and lengthy induction. Surprisingly, the rapid induction was not maintained over time, as expected, until the final recovery of the steady state levels; instead, it decreased suddenly. This effect was also reflected in the cardenolide content.

The P5βR/P5βR2 genes are responsible for the first step of the cardenolide-specific biosynthetic pathway, that is, the reduction of progesterone to a 5β-pregnane. Although the genes catalyzing the following steps towards genin formation have not been isolated and characterized, the correlation between P5βR2 expression and cardenolide formation demonstrates that this gene plays a pacemaker role in the synthesis of these natural products. These findings confirm the observations reported previously by Gärtner & Seitz (1993) on the activity of the native enzyme, who suggested that progesterone reduction was a limiting step in the cardenolide pathway.

Morales et al. (1993) reported that moderate salinity conditions led to raised cardenolide levels in D. purpurea. However, Paranhos et al. (1999) demonstrated that exogenous H₂O_2, or superoxide dismutase, stimulated cardenolide production in cell cultures of D. thapsi, whereas the addition of catalase markedly reduced it. These authors suggested a connection between salinity or H₂O₂ and cardenolide formation, which is now supported by the inducible expression of P5βR2 observed in Digitalis plants on external treatments with NaCl and H₂O₂.

Wounded foxglove plants showed significant variations in their cardenolide composition, alterations that not only affected the quantity of these metabolites, as discussed earlier, but also their qualitative profiles. Cardenolide analyses revealed that, after wounding, plants contained more nonpolar cardiac glycosides, which was basically reflected by an increase in the DM concentration and by a lower accumulation of digitoxin. The levels of the corresponding genin, digitoxigenin, remained basically constant after wounding in comparison with the control. We suggest that the absence of a significant variation in digitoxigenin content is a result of its glycosylation reaction catalyzed by specific 3β-hydroxy sterol glycosyltransferases, as the major product of such a reaction would be DM. Glycosylation enables the stabilization and storage of potent toxic metabolites in high concentrations, whilst simultaneously allowing the plant to release highly deterrent chemicals on decompartmentalization and exposure to specific glycosidases (Vogt & Jones, 2000). The involvement of glycosyltransferases in the response of plants to biotic and abiotic stress, including wounding, has been reported (O’Donnell et al., 1998; Gachon et al., 2005; Madina et al., 2007). Nonetheless, we were unable to substantiate such a correlation between DM accumulation and increased glycosyltransferase activity, induced by wounding, because the genes encoding glycosyltransferases in Digitalis are unknown.

In agreement with Fordyce & Malcolm (2000), these results may be interpreted as an increase in the toxicity of plants, as the more lipophilic the compounds, the easier they pass across biological membranes. Therefore, qualitative variations in secondary metabolites may play a role in plant defense which is as relevant as that of the quantitative changes observed in response to herbivore attacks. Interestingly, the CGH enzyme deglucosylating primary cardenolides in Digitalis may also be an important factor in the defense of these species against predators, because the secondary cardenolides formed are better resorbed and more toxic than their primary counterparts (Framm et al., 2000).

Together, these results indicate that P5βR2 is a critical component for the chemical defense of foxglove plants against herbivores, through cardenolide accumulation, in association with ethylene and H₂O₂ signaling. Currently, there are no data available on additional genes that may be involved in cardenolide biosynthesis, and therefore the question is whether the reduction of progesterone is the only rate-limiting step or whether other enzymes coming later in the pathway have a regulatory role.

Acknowledgements

We are grateful to the SCSIE of the University of Valencia for technical assistance and to the Plataforma Bioinformática de Andalucía for computational and technical support. The authors thank two anonymous referees for stimulating comments on an earlier version of this article. This work was supported by the Generalitat Valenciana (Project GV05/127), Universidad de Valencia (AE-20070215) and Ministerio de Ciencia e Innovación (BFU2008-01420/BMC). IG received financial support from a Ramón y Cajal contract from the Spanish Ministerio de Educación y Ciencia.