Cloning and expression of two novel aldo-keto reductases
from
Digitalis purpurea
leaves
Isabel Gavidia
1,2
, Pedro Pe
´
rez-Bermu
´
dez
2
and H. Ulrich Seitz
1
1
Center of Plant Molecular Biology (ZMBP), University of Tu
¨
bingen, Germany;
2
Department of Plant Biology, University of
Valencia, Spain
The aldo-keto reductase (AKR) superfamily comprises
proteins that catalyse mainly the reduction of carbonyl
groups or carbon–carbon double bonds of a wide variety of
substrates including steroids. Such types of reactions have
been proposed to occur in the biosynthetic pathway of the
cardiac glycosides produced by Digitalis plants. Two cDNAs
encoding leaf-specific AKR proteins (DpAR1 and DpAR2)
were isolated from a D. purpurea cDNA library using the rat
D
4

Many of these natural products have been shown to have
important ecological functions, comprising resistance
against diseases (phytoalexins) and herbivore (proteinase
inhibitors, bitter and toxic deterrents, etc.). Besides this,
plant secondary metabolism is the source for many fine
chemicals such as drugs, dyes, flavours and fragrances, all of
increasing commercial importance. Therefore, the possibil-
ities to alter the production of secondary metabolites are of
great interest, but the limited knowledge of the biosynthetic
routes, often based only on feeding experiments and/
or theoretical considerations, is a major constraint in this
field [3].
One group of natural products of major interest in the
pharmaceutical industry is cardiac glycosides from Digitalis
species, as they are widely prescribed for the treatment of
congestive heart failure. Cardiac glycosides possess a basic
skeleton, a steroid genin, namely digitoxigenin, digoxigenin
or gitoxigenin. Different studies using labelled and unla-
belled precursors have led to a hypothetical pathway for
cardenolide biosynthesis, but knowledge about the forma-
tion of the aglycon is not well established. The first steps of
this route basically resemble those of cholesterol metabolism
towards steroid hormones in animals. Upstream of digit-
oxigenin, only four reactions have been described: the
transformation of cholesterol to pregnenolone [4], prog-
esterone formation from pregnenolone [5], the sequential
reductions of progesterone to 5b-pregnane-3,20-dione [6]
and 5b-pregnan-3b-ol-20-one [7]. In animal tissues all of
these reactions of the steroid metabolism, except the
cholesterol side-chain-cleaving reaction, are catalysed by

rtner et al. [9] and the
subsequent amino acid sequencing.
The second strategy is based on the use of orthologous
genes for screening a cDNA library of D. purpurea.Assuch
a type of steroid stereospecific enzyme has never been cloned
in plants, and considering that some aspects of the steroid
metabolism in higher plants cannot be separated from
corresponding ones in animals, the cDNA encoding D
4
-3-
ketosteroid 5b-reductase of rat liver [10] was used as a
probe. The results obtained in this second experimental
approach are described in the present paper, which reports
on the cloning and expression of two AKR genes from
D. purpurea. Both proteins reduce the ketone group of
steroid structures but they are not active on the D
4
-double
bond of the steroids assayed. It is worth noting that this is
the first report for such activity on steroids from a plant
AKR enzyme.
MATERIALS AND METHODS
Plant materials
Shoot cultures of D. purpurea were established as described
previously [7]. Every 3 weeks newly developed shoots were
transferred to fresh liquid nutrient medium [11] supple-
mented with 3% glucose, 1 mgÆL
)1
indoleacetic acid and
2mgÆL

4
-3-ketosteroid 5b-reductase
cDNA as a probe. Nylon filter lifts were prehybridized and
hybridized at 42 °Cin330m
M
sodium phosphate buffer
(pH 7), 7% SDS, 1 m
M
EDTA and 1% BSA. The positive
clones were isolated, their cDNA inserts in vivo excised, then
subcloned into pBluescript SK(–).
DNA sequence analysis
Restriction analysis was used to classify the clones.
cDNA clones were subjected to nucleotide sequencing
by the dideoxy chain termination method, using a DNA
sequencing kit (PE Biosystems) on an ABI 310 Genetic
Analyser (PerkinElmer). Complete nucleotide sequences
were determined for both strands of the cDNAs and
analysed by the
DNASTAR
program package (Lasergene)
and
CLUSTALW
.
Plant treatments
To determine the effects of several stress conditions on gene
expression, D. purpurea plants were grown in a greenhouse
for 4 months. For mechanical wounding experiments, holes
of 1 mm diameter were made across the lamina, which
effectively damaged approximately 5% of the leaf area.

).
RNA and DNA gel blots were prehybridized for 3 h at
50 °C in a solution containing 330 m
M
sodium phosphate
buffer (pH 7), 7% SDS, 1% BSA, 1 m
M
EDTA. Hybrid-
ization was done overnight at 65 °C with a random primed
32
P-labelled cDNA probe (SmaI fragment of 900 bp from
DpAR1). The membranes were finally washed in 2 · NaCl/
Cit, 0.1% SDS for 20 min at 65 °C. Autoradiography of the
filters was obtained on X-OMAT AR films (Kodak) using
an intensifying screen at )80 °C.
Expression and purification of recombinant
DpAR proteins
ThecDNAswereclonedasaSphI/BglII fragment into
pQE70 QIAexpress vector (Qiagen). Both recombinant
plasmids were transfected into Escherichia coli strain M15/
PREP4. Cells were grown in Luria–Bertani media supple-
mented with 100 lgÆmL
)1
ampicillin and 25 lgÆmL
)1
kanamycin at 37 °C. Gene expression was induced by the
addition of isopropyl thio-b-
D
-galactoside to a final con-
centration of 1 m

Enzyme assays
Enzyme activity was determined in a 1-mL reaction mixture
containing 0.1
M
sodium phosphate buffer (pH 7.0);
150 l
M
NADPH, NADH, NADP or NAD; 10 m
M
DL
-glyceraldehyde,
D
-glucose or
D
-fructose; 10 l
M
prog-
esterone, 5b-pregnan-3,20-dione, 17a-hydroxiprogesterone
or 5b-pregnan-3b-ol-20-one. The reaction was initiated by
the addition of the protein, and monitored at 25 °Cusinga
Uvicon 930 spectrophotometer (Kontron Instruments,
Germany). The activity was determined by measuring
NADPH, NADH oxidation or NADP, NAD reduction
from the decrease or increase in absorbance at 340 nm,
respectively. Steroids were dissolved in ethanol, which did
not exceed 5% of the total volume. The appropriate blank
was subtracted from each determination to correct nonspe-
cific oxidation or reduction of cosubstrate. One unit of
enzyme activity was defined as the amount that oxidized
1 lmol NAD(P)HÆmin

nucleotide sequences of DpAR1 and DpAR2 have been
submitted to the EMBL database and are available under
accession numbers AJ309822 and AJ309823, respectively.
DpAR1 and DpAR2 contain 948 bp long ORFs encoding
315 amino acids of a calculated molecular mass 34 898 and
34 883 Da, respectively. Their nucleotide sequences exhibit
97.8% identity, and their amino-acid sequences show a
98.4% identity.
The sequence comparison analysis revealed the significant
homology of DpAR1 and DpAR2 to the AKR protein
superfamily. Comparison of DpAR1 and DpAR2 amino-
acid sequences with those of plants (Fig. 1) revealed that
these proteins present 80, 73, 70 and 68% identity with four
Arabidopsis thaliana clones (accession numbers AAC23647,
AAD32792, AAC23646 and CAB88350, respectively) and
67% with the alfalfa aldose/aldehyde reductase (accession
number X97606). Furthermore, we found a 45–47%
conservation in amino-acid residues with AKR4 proteins
from plants and 40–42% with AKR1 proteins from human
and animals.
Out of the four Arabidopsis clones, only one (CAB88350)
has been appointed as a putative AKR, whereas the other
three are postulated to be alcohol dehydrogenases. Never-
theless, the high degree of homology of their amino-acid
sequences and those of the alfalfa and D. purpurea ARs
suggests that all these proteins belong to a plant AKR
subfamily. Our results are in agreement with the conclusions
of Oberschall et al. [16] based on the comparison of only
two Arabidopsis clones with the AKR sequence of Medicago
sativa.

in NAD(P)H binding. The residues K256 and S257 are part
of a typical AKR motif (IPKS) having cosubstrate binding
functions [21]. Although this motif is highly conserved, all
residues are not invariant, as happens within the subfamily
proposed (see Fig. 1) where the residue I changes to L.
Genomic organization of
D. purpurea
AR genes
The molecular organization of the AR genes in D. purpurea
was determined by Southern blot analysis of genomic DNA
digested with EcoRI, BamHI and HindIII. There were no
2844 I. Gavidia et al. (Eur. J. Biochem. 269) Ó FEBS 2002
BamHI or HindIII restriction sites in any of the DpAR
cDNAs, and only one EcoRI site in the DpAR2 clone. The
900-bp cDNA fragment of DpAR1 was used as a probe.
After washing the blotting membrane under high-stringent
conditions, five bands were detected for cuts by BamHI or
HindIII enzymes while up to 10 bands were found when
genomic DNA was digested with EcoRI (Fig. 3). These
results indicate that a small multigene family of at most five
genes is encoding ARs in the genome of D. purpurea.
Heterologous expression of DpARs in
E. coli
The cDNAs of DpAR1 and DpAR2 were over-expressed in
E. coli as fusion proteins with His-tag (pQAR1 and
pQAR2). The recombinant DpARs were purified by affinity
chromatography from the extracts of bacteria transformed
with pQAR1 or pQAR2, using a His-binding resin column.
The purified proteins were visualized as a single band of
35 kDa after SDS/PAGE. To test the cofactor specificity of

substrates including steroids.
Preliminary results indicated that these enzymes work by
reducing progesterone and 17a-hydroxyprogesterone; mol-
ecules having 20-one and D
4
-3-one structures, which are
susceptible for such reduction. TLC analysis of the steroids
extracted from the enzymatic reactions, using progesterone
as a substrate, showed a lack of fluorescence with respect to
the control (Fig. 4); this could be due to the reduction of the
D
4
-double bond and/or the ketone group at position 3. To
determine the specific function of DpAR recombinant
enzymes, different steroids lacking one or more of such
structures were assayed. 5b-pregnan-3,20-dione (having
3- and 20-one structures) served as substrate with a reaction
rate almost identical to that obtained for progesterone.
Fig. 1. Alignment of the amino-acid sequences
of proteins within a closely related AKR family.
Sequences aligned are DpAR1 and DpAR2,
D. purpurea (this study); AAC23647,
AAD32792, CAB88350 and AAC2346,
Arabidopsis thaliana;Medicago,M. sativa
[16]. The amino-acid residues identical to the
DpAR1 sequence are indicated by dots. Gaps
are introduced to optimize the alignment.
Ó FEBS 2002 Aldo-keto reductases in D. purpurea leaves (Eur. J. Biochem. 269) 2845
However, 5b-pregnan-3b-ol-20-one (having only 20-one
structure) showed lower specific activity (Table 1). Com-

Glucose 1.80 ± 0.09 0.57 ± 0.05
Fructose 1.83 ± 0.15 0.60 ± 0.05
Progesterone 1.77 ± 0.16 1.49 ± 0.09
5b-Pregnane-3,20-dione 1.70 ± 0.10 1.45 ± 0.11
5b-Pregnan-3b-ol-20-one 1.32 ± 0.08 1.04 ± 0.12
Fig. 4. TLC analysis of steroids visualized under UV-light. Lane 1,
authentic progesterone (P); lane 2, reaction mixture with DpAR1
recombinant protein and NADH as cosubstrate; lane 3, reaction
mixture without NADH; lane 4, reaction mixture without protein.
2846 I. Gavidia et al. (Eur. J. Biochem. 269) Ó FEBS 2002
of the steroids assayed. It is worth noting that this is the first
report for such steroid activity from a plant AKR enzyme.
Expression of DpAR genes
The size of the DpAR transcripts and their expression
profile were determined by Northern hybridization analysis
of total RNA. As shown in Fig. 5, a single mRNA species
with a size of  1.4 kb was detected with the 900 bp
DpAR1 cDNA probe. The organ-specific expression of
DpAR genes in mature D. purpurea plants (1 year old) is
shown in Fig. 5A. A highly specific expression profile was
obtained, as the hybridization signal was restricted to the
leaf blade. No signals were detected in the petiole, stem or
roots even after over-exposure of the film. The transcription
level was also examined during in vitro development of
D. purpurea shoot cultures, leaf samples being taken at
different time points. DpAR expression slightly increased
along the culture time course, although at the end of the
experiment (3 months) the transcription level of the in vitro
plants was clearly weaker than in mature field plants
(Fig. 5B).

catalysing the reduction of the D
4
-double bond to give A/
B-cis conformation [10]. A similar reaction, reduction of
progesterone to 5b-pregnane-3,20-dione, catalysed by prog-
esterone 5b-reductase, has been considered as the stereo-
specific starting point of the cardenolide pathway leading to
digitoxigenin [22]. Thus, a heterologous AKR clone, D
4
-3-
ketosteroid 5b-reductase [10] from rat, has been used for
screening a D. purpurea cDNA library. Although the aldo
and keto groups of the substrate are not involved chemically
in the reaction, this enzyme is classified as a member of the
AKR superfamily because it shares  50% homology and
the typical signatures with other members of this family,
including the ARs.
Following this cloning strategy, we isolated and
sequenced two full-length cDNAs from D. purpurea leaves
that encode DpAR1 and DpAR2, two new members of the
AKR superfamily; specifically, the amino-acid sequences of
DpARs show relatively high levels of similarity to mam-
malian ARs. The highest identities were obtained with
several Arabidopsis proteins of unknown function and the
aldose-aldehyde reductase of M. sativa [16]. Lower levels of
similarity, within the plant AKRs, were found with the AR
proteins from the monocotyledoneous Avena fatua [23],
Hordeum vulgare [24], Bromus inermis [25] and Xerophyta
viscosa [21].
Fig. 5. Expression of DpAR genes. For all Northern analysis, 20 lg

not been linked to steroid metabolism.
As has been reported for both animal and plant AKRs,
the corresponding protein expressed in bacteria has the
same properties as the in vivo protein [32]. Purification of the
recombinant DpARs from E. coli allowed us to show that
these enzymes are capable of reacting with sugars and
steroids. The typical aldose substrates
DL
-glyceraldehyde
and
D
-glucose, as well as the ketose
D
-fructose, were reduced
in the presence of NADH by both enzymes with similar
activities. Two important differences have been observed
when comparing DpARs with the AR from barley and
alfalfa, as these proteins use NADPH as cosubstrate, and
their activities with glyceraldehyde were clearly higher than
with glucose. When reacting with steroids, DpAR1 and
DpAR2 cannot reduce the carbon–carbon double bond
in D
4
-3-ketosteroids, but have been shown to reduce both
3- and 20-ketosteroids. Thus, DpARs are ketosteroid
reductases instead of D
4
-3-ketosteroid 5b-reductases. As
inferred from the results of their enzymatic activities,
DpAR1 and DpAR2 may be two isoforms with different

DpARs, we found slight differences between their substrate
specificity, which may be related to the variation of certain
amino acids, and it can be also assumed that the natural
substrates for these DpARs have not been detected in our
system. A problem commonly connected with AKR
enzymes is their broad substrate specificity, which makes
it difficult to determine the physiologically used substrate(s)
and consequently the physiological role(s) of a particular
enzyme of this family. Thus, in our case both enzymes
function not only as typical ARs, but also as ketosteroid
reductases; this suggests their involvement in steroid meta-
bolism. DpAR1 and DpAR2 show enzymatic activity with
some intermediate products of the pregnane metabolism.
Accordingly, both proteins may participate in the formation
of a-orb-pregnane derivatives. The latter case would imply
their involvement in the pathway of cardenolide biosynthe-
sis as b-configured pregnanes are the putative precursors of
these natural products.
Northern analysis revealed the tissue-specific expression
of DpAR genes in D. purpurea plants, showing a specific
signalwhichisrestrictedtoleaves.Furthermore,the
transcription level increased with plant development as
mature leaves exhibited higher expression levels than those
of young plantlets. The lack of specific probes for each
gene did not permit determination of whether both
DpAR genes exhibit differential expression associated to
the plant developmental stage. These results allow us to
establish interesting correlations between the enzymatic
activity on steroids, the organ-specific and developmen-
tally regulated expression of the genes, and the specific

stress and heat-shock treatment. The response of plant AR
enzymes to a wide range of stresses was also observed in
M. sativa [16], wherein a physiological role in plant defence
has been attributed; the authors suggested that such
resistance might primarily be due to detoxification of toxic
aldehydes. The stimulation of AR synthesis under stress
conditions points to a physiological role of these enzymes in
plants exposed to environmental stresses.
In conclusion, we have isolated two cDNA clones and
determined the primary structure of two AKRs from
D. purpurea. These proteins share considerable similarities
not only with plant ARs but also with the mammalian
AKRs having a role in steroid metabolism. This observation
is the first report that biologically active steroids are
substrates for plant AKRs. These results, besides others
mentioned above, suggest that DpARs are involved in
cardenolide biosynthesis. Nevertheless, the limited know-
ledge on the intermediates and enzymes of this biosynthetic
pathway, besides the broad substrate specificity of these
AKRs, are major restrictions to elucidate the physiological
role of these proteins in D. purpurea. Work is underway to
determine the precise role of DpARs in plant steroid
metabolism in general and in cardenolide biosynthesis in
particular. More experiments will be necessary to determine
their link with stress tolerance. This knowledge would be of
great importance not only for these plant processes but also
for a comparison of multifunctional roles of AKRs in plants
and mammals.
ACKNOWLEDGEMENTS
The authors thank Dr Temp. M. Noshiro (Hiroshima University) for

rtner, D.E., Wendroth, S. & Seitz, H.U. (1990) A stereospecific
enzyme of the putative biosynthetic pathway of cardenolides.
Characterisation of a progesterone 5b-reductase from leaves of
Digitalis purpurea L. FEBS Lett. 271, 239–242.
Ó FEBS 2002 Aldo-keto reductases in D. purpurea leaves (Eur. J. Biochem. 269) 2849
7. Ga
¨
rtner, D.E. & Seitz, H.U. (1993) Enzyme activities in carde-
nolide-accumulating, mixotrophic shoot cultures of Digitalis pur-
purea L. J. Plant Physiol. 141, 269–275.
8. Penning, T.M., Ma, H. & Jez, J.M. (2001) Engineering steroid
hormone specificity into aldo-keto reductases. Chem. Biol. Inter-
act. 130, 659–671.
9. Ga
¨
rtner, D.E., Keilholz, W. & Seitz, H.U. (1994) Purification,
characterization and partial peptide microsequencing of prog-
esterone 5b-reductase from shoot cultures of Digitalis purpurea.
Eur. J. Biochem. 225, 1125–1132.
10.Onishi,Y.,Noshiro,M.,Shimosato,T.&Okuda,K.(1991)
Molecular cloning and sequence analysis of cDNA encoding D
4
-3-
ketosteroid 5b-reductase of rat liver. FEBS Lett. 2, 215–218.
11. Murashige, T. & Skoog, F. (1962) A revised medium for rapid
growth and bioassay with tobacco tissue culture. Physiol. Plant.
15, 473–496.
12. Steimle, D.E., Schnitzler, J.P. & Seitz, H.U. (1994) Modes of ex-
pression of phenylalanine ammonia-lyase and chalcone synthase
in elicitor-treated carrot cell cultures. Acta Hort. 381, 206–209.

(1995) Mechanism of human aldehyde reductase: characterization
oftheactivesitepocket.Biochemistry 34, 11264–11275.
20. Jez, J.M. & Penning, T.M. (1998) Engineering steroid 5b-reduc-
tase activity into rat liver 3a-hydroxysteroid dehydrogenase. Bio-
chemistry 37, 9695–9703.
21. Mundree, S.G., Whittaker, A., Thomson, J.A. & Farrant, J.M.
(2000) An aldose reductase homolog from the resurrection plant
Xerophyta viscosa Baker. Planta 211, 693–700.
22. Seitz, H.U. & Ga
¨
rtner, D.E. (1994) Enzymes in cardenolide-
accumulating shoot cultures of Digitalis purpurea L. Plant Cell
Tiss. Org. Cult. 38, 337–344.
23. Li, B. & Foley, M.E. (1995) Cloning and characterization of dif-
ferentially expressed genes in imbibed dormant and afterripened
Avena fatua embryos. Plant Mol. Biol. 29, 823–831.
24. Bartels, D., Engelhardt, K., Roncarati, R., Schneider, K.,
Rotter, M. & Salamini, F. (1991) An ABA and GA modulated
gene expressed in the barley embryo encodes an aldose reductase
related protein. EMBO J. 10, 1037–1043.
25. Lee, S.P. & Chen, T.H.H. (1993) Molecular cloning of abscisic
acid-responsive mRNAs expressed during the induction of freez-
ing tolerance in bromegrass (Bromus inermis Leyss) suspension
culture. Plant Physiol. 101, 1089–1096.
26. Garcı
´
a-Pe
´
rez,A.,Martin,B.,Murphy,H.R.,Uchida,S.,
Murer,H.,Cowley,B.D.,Handler,J.S.&Burg,M.B.(1989)

diabetic complications. Science 257, 81–84.
35. Gui, T., Tanimoto, T., Kokai, Y. & Nishimura, C. (1995) Presence
of a closely related subgroup in the aldo-keto reductase family of
the mouse. Eur. J. Biochem. 227, 448–453.
36. Malcolm, S. & Zalucki, M. (1996) Milkweed latex and cardenolide
induction may resolve the lethal plant defence paradox. Entomol.
Exp. Appl. 80, 193–196.
37. Saitou, N. & Nei, M. (1987) The neighbor-joining method: a new
method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4,
406–425.
2850 I. Gavidia et al. (Eur. J. Biochem. 269) Ó FEBS 2002