Insights into the reaction mechanism of glycosyl hydrolase family 49
Site-directed mutagenesis and substrate preference of isopullulanase
Hiromi Akeboshi
1
, Takashi Tonozuka
1
, Takaaki Furukawa
1
, Kazuhiro Ichikawa
1
, Hiroyoshi Aoki
1,2
,
Akiko Shimonishi
1
, Atsushi Nishikawa
1
and Yoshiyuki Sakano
1
1
Department of Applied Biological Science, Tokyo University of Agriculture and Technology, Fuchu, Tokyo, Japan;
2
Fuence Co., Shibuya, Tokyo, Japan
Aspergillus n iger isopullulanase (IPU) is the only pullulan-
hydrolase in glycosyl hydrolase (GH) family 49 and does not
hydrolyse d extran at all, while all o ther GH family 49
enzymes a re dextran-hydrolysing enzymes. T o i nvestigate
the common catalytic mechanism of GH family 49 enzymes,
nine mutants were prepared to replace residues conserved
among GH family 49 (four Trp, three A sp and two Glu).
Homology modelling of I PU was also carried out based o n

(1fi6)-Glc-a-(1fi4)-Glc) s tructure, and cleaves the a-1,4-
glucosidic linkage in the panose motif [1,2]. Enzymes that
hydrolyse specific sites of pullulan can be classified into the
following three types (schematic action patterns of these
enzymes have been illustrated previously [2]). (a) Pullulanase
(EC 3 .2.1.41), which hyd rolyses a-1,6-glucosidic linkages t o
produce maltotriose [3]; (b) Thermoactinomyces vulgaris
R-47 a-amylase (TVA, EC 3.2.1.1) [4] and neopullulanase
(EC 3.2.1.135) [5], which hydrolyse a-1,4-glucosidic linkages
to produce panose; and (c) IPU, which h ydrolyses the other
a-1,4-glucosidic linkages to produce isopanose. Except for
IPU, these enzymes are c lassified into glycosyl h ydrolase
(GH) family 13, known as the a-amylase family (reviewed in
[6–8]). In contrast, IPU is the sole enzyme classified into GH
family 49 [2,9,10] among these pullulan-hydrolases, and no
homology between IPU a nd a-amylase family e nzymes
has been found ( />index.html).
Interestingly, IPU does not hydrolyse dextran at all, while
all o ther GH family 49 enzymes are dextran-hydrolysing
enzymes, such as endo-dextranase (EC 3.2.1.11) [11–14] and
isomaltotrio-dextranase (EC 3.2.1.95) [15]. We have repor-
ted the molecular cloning of IPU, and indicated that seven
highly co nserved regions are found among the p rimary
structures of these dextran-hydrolases and IPU [2]. The
expression systems of IPU have been constructed with
eukaryotic hosts Aspergillus oryzae and Pichia pastoris
[2,16]. Recently, a three-dimensional structure of GH family
49 dextranase (D ex49A), wh ich shows a 38% s equence
identity with IPU, has b een reported, and the catalytic
domain folds into a right-handed parallel b-helix [17].

chemical modification experiment IPU was inactivated by
N-bromosuccinimide, i ndicating that a Trp residue is
required f or IPU activity [27]. Therefore, several Asp and/
or Glu, and Trp residues located in the conserved regions
are predicted to be indispensable for IPU a nd the other GH
family 49 enzymes. Here we determined the residues that are
essential for the catalytic activity of IPU, and also investi-
gated some detailed properties of this enzyme. The results
indicated that the functionally important residues o f GH
family 49 enzymes are conserved in the negatively numbered
subsites, and the substrate preference of the negatively
numbered subsites of IPU also resembles that of GH family
49 dextranases.
Materials and methods
Host strains and media
Escherichia coli JM109 was used for the plasmid con-
structions. P. pastoris GS115 (Invitrogen) was used for the
heterologous expression of IPU. The Luria–Bertani
medium for E. coli, and the yeast peptone dextrose
medium and buffered minimum methanol medium
(BMM) for P. pastoris were prepared according t o t he
manufacturer’s r ecommendations. Y east peptone glycerol
medium (YPGY: 1% yeast extract, 2% peptone, and 1%
glycerol) was prepared for the propagat ion of recombinant
P. pastoris strain for t he expression of enzymes. All
cultivation was done at 37 °CforE. coli and 30 °Cfor
P. pastoris.
Mutant constructions
All c loning procedures were carried out by applying
standard molecular b iological t echniques [28]. Transforma-

clear supernatant of cultured BMM (crude IPU) was then
obtained by c entrifugation ( 5000 g for 1 5 min). During the
cultivation in BMM, 500 lL of methanol was added every
24 h for the maintenance of 0.5% (v/v) methanol as the
carbon source and i nducer. The expressi on of IPU mutants
was performed using the same procedure as for the wild-
type IPU.
Substrates
Pullulan w as obtained from H ayashibara, Japan. Panose
[30], IMTG (4
1
-a-isomaltotriosylglucose, also called
6
2
-a-isomaltosylmaltose; Glc-a-(1fi6)-Glc-a-(1fi6)-Glc-a-
(1fi4)-Glc) [ 31], and MM (6
2
-a-malto sylmaltose; Glc-a-
(1fi4)-Glc-a-(1fi6)-Glc-a-(1fi4)-Glc) [32] were prepared
as described. Dextran T-2000 was from Amersham.
Concentrations of the substrates dissolved in 50 m
M
acetate buffer (pH 3.5) were measured by a modified
phenol/sulfuric acid method [33], using glucose as the
standard.
Table 1. Primers used for the mutant constructions. For PCR muta-
genesis (W21F, W32F, W240F and W402F), the same sequences on the
opposite strand were also used as described in Materials and methods.
Lower c ase letters indicate the nucleotide m utations. To facilitat e the
selection of mutant clones, silent mutations were made to introduce

Assays of IPU activity and protein concentration
The p ullulan-hydrolysing activity of IPU was evaluated as
described previously [34]. The activities for panose, IMTG
and MM were measured as follows. A re action mixture
consisting of a desired substrate (32 m
M
)andIPUin
40 m
M
acetate buffer (pH 3.5) was incubated at 40 °Cfor
30 min, and the reaction was terminated by t he addition
of an equal v olume o f 0.1
M
Na
2
CO
3
. T he amount of
glucose p roduced by IPU was assayed with G lucose CII
(Wako Pure Chemical Co., Osaka, Japan) [35,36] using a
Bio-Rad550 Microplate reader. To determine the kinetic
parameters of wild-type and W402F IPU for p anose,
mixtures consisting of 4 lgÆmL
)1
enzyme and from 43 to
216 m
M
substrate, and 20 lgÆmL
)1
enzyme and 64 to

(PDB IDs, 1OGM and 1OGO, respectively) [17]. The
primary structure of mature IPU (residues 20– 564) was
submitted for automatic modelling on t he Swiss-Model
server ( asy.org/) [ 38] u sing the first
approach mode, a nd a m odel consisting of residues 2 5–540,
which is based on the structure of 1OGM, was obtained. To
determine the potential catalytic site, this mo del was
superimposed on 1OGO using the program
DEEPVIEW
[38],
and a glucosyl units in subsites +1 and +2 w ere placed i n
the m odel. The figure was generated using the programs
RASMOL
[39],
RASTOP
(einfinity.org/rastop/),
MOLSCRIPT
[40] and
RASTER
3
D
[41].
Polarimetric assays
Polarimetric measurements of IPU, glucoamylase from
Rhizopus niveus (Seikagaku K ogyo, Japan; a nomer-invert-
ing enzyme; EC 3.2.1.3), and a-glucosidase from Bacillus sp.
(Toyobo, Japan; an omer-retaining enzyme; EC 3 .2.1.20)
were compared. An enzyme solution (equivalent to
1.2 UÆmL
)1

specific activity was also h igher t han t hose o f IPU from
original A. niger and heterologously expressed I PU fr om
A. oryzae (27 a nd 38ÆUmg
)1
, r espectively) [7,27]. The
mutatedIPUswerepurifiedwiththesameprocedureas
wild-type IPU.
Properties of Trp-mutated enzymes
Previous experiments indicated that some T rp residues are
essential f or IPU a ctivity [27]. Four Trp residues, conserved
in the seven regions of GH family 49 (Fig. 1), are replaced
by Phe (W31F, W32F, W240F and W402F). The relative
activity of mutated enzymes towards pullulan and panose
are shown in Table 2. The W 402F enzyme lost the activity
for pullulan ( 0.4% of the wild-type IPU) and the activity
for p anose w as almost undetectable ( 0.1%) under the
given conditions. The W31F enzyme had only 38% activity
for pullulan, but the a ctivity for panose w as 1.4-fold higher
than that of wild-type enzymes. The activities of W32F
and W 240F were similar (90–160%) to that o f wild-type
enzyme.
As t he a ctivity o f W402F drastically decr eased, it s action
pattern was investigated using TLC. T he W402F enzyme
liberated isopanose from pullulan, and isomaltose and
glucose from panose, and the action patterns of the wild-
type and the W402F enzymes were almost identical (Fig. 2).
The kinetic study for W402F towards panose showed that
the K
m
value was sixfold higher and the k

uct, isomaltose, has been reported (PDB ID, 1OGO) [17].
The identity between the primary structures of Dex49A a nd
IPU is 38%, and a t hree-dimensional structure of IPU was
modelled based on the structure of Dex49A using the Swiss-
Model server. IPU consists of a signal sequence (residues
1–19) and a mature part (residues 20–564), and the model is
composed of residues 25–540. The overall structure of the
model of IPU and the mutated residues in this study are
shown in Fig. 3. To elucidate the mechanism of the
substrate recognition of IPU, two glucosyl units (Glc +1
and +2, respectively) were forced to be placed based on the
position o f isomaltose bound in the subsites +1 and + 2 of
Dex49A, although IPU does not produce isomaltose.
IPU was predicted to consist of two domains, N-terminal
domain (residues 25–189) and b-h elical domain ( residues
190–540). Asp353, Glu356, Asp372, Asp373, and Trp402,
whose substitutions resulted in the reduction of the activity
for both pullulan an d panose, were predicted to be l ocated
in potential subsites )1and)2 ( a d etailed description is
given i n the next section). T rp31 and Glu273, whose
Fig. 1. Conserved regions of GH family 49 e nzymes. Identical amino acid residues are shown in white on black, and conserved Trp, Asp and Glu
residues are indicated b y asterisks. PMDEX, Penicillium minioluteum dextranase [12]; DEX49A, Penicillium minioluteum dextranase isoform [13];
PFDEXA, Penicillium funiculosum dextranase (DDBJ/EMBL/GenBank No. AJ272066); AGTDEX1 and 2, Arthrobacter globiformis T-3044
endodextranase 1 an d 2 [1 4]; AGCDEX, Arthrobacter sp. CB-8 d extranase [11]; I MTD, Brevibacterium fuscum var. dextranlyticum isomaltotrio-
dextranase [15].
Table 2. Relative activities of wild-type and mutant IPUs for pullulan
and panose. Activities for 0.4% (w/v) pullulan and 32 m
M
panose were
measured. ND, Not d etected.

) k
0
(s
)1
) k
0
/K
m
(m
M
)1
Æs
)1
)
Wild-type 160 ± 3.8 180 ± 2.2 1.13 ± 0.04
W402F 920 ± 140 0.36 ± 0.036 (3.9 ± 1.0) · 10
)4
Ó FEBS 2004 Reaction mechanism of isopullulanase (Eur. J. Biochem. 271) 4423
substitutions caused a decrease in the activity for pullulan
(38 and 45%, r espectively) but not significant for panose
(140 and 74%, respectively), are located r elatively far from
the potential catalytic site, and the side chains were
predicted t o o rient t o the interface b etween N-terminal
and b-helical domains. A structural homology search for the
N-terminal domain (residues 25–189) was also carried out
using the Dali server [44]. N umerous proteins containing an
immunoglobulin-like fold were listed, and among glycosyl
hydrolases, domain N of a pullulan-hydrolysing enzyme
from Thermoactinomyces vulgaris, TVA II (PDB ID, 1BVZ;
Z score of 3.2) [45] was a solution in the Dali result. It is

aromatic and charged residues Arg297–322, Asn323–348,
Asp326–351, Tyr358–381, Lys376–399, Tyr378–401,
Tyr379–402, and Tyr440–463, are conserved in the negat-
ively numbered subsites. On the other hand, residues located
in the positively numbered s ubsites are relatively not
conserved. The report o f Dex49A shows an illustration
where seven amino acid residues, Asp86, Tyr303, Lys315,
Asp395, Asn417, Lys447, and Glu449, interact with Glc +1
and +2 [17] (Fig. 4B). Only two of these residues,
Tyr278(IPU)-303(Dex49A), and Asp372–395, both o f
which interact with Glc +1, are conserved between IPU
and Dex49A. The position equivalent to Lys315 o f Dex49A
is identified as Gly290 of IPU, which may enable IPU to
incorporate the a-(1fi4)-linked glucose un its. In addtion, in
Dex49A, Phe373 protrudes to the active cleft, which appears
to restrict the c onformation of the s ubstrate and accom-
modate only the a-(1fi6)-linked glucose units. In IPU, the
position e quivalent to Phe373 of Dex49A is identified as
Gly350, which allows IPU to have a relatively wide cleft,
thus it seems to b e possible that both a-(1fi4)-linked and a-
(1fi6)-linked glucose units enter the active cleft of I PU.
However, residues corresponding to Asn417 and Glu449 o f
Dex49A are v irtually lack ing in I PU because positions
equivalent to Asn417 and Glu449 of Dex49A are ident ified
as Val39 4 and Gly426 of IPU, respectively. Therefore, even
if the a-( 1fi6)-linked glucose units enter to the active cleft of
IPU as shown in Fig. 4 A, it would be impossible for the
substrate to be retained in the cleft. In the model of IPU,
several aromatic and charged residues, Trp277, Tyr349, and
Asp371, are p resent in t he v icinity of G ly350, and c ould b e

In the case of inverting enzymes, a single displacement
mechanism has been proposed [6,17,19–21]. In this mech-
anism, two catalytic residues function as a general acid
(donating a p roton) and a general base ( activating the
nucleophilic water m olecule) in the first step of the reaction.
A carbonium ion intermediate subsequently forms, and is
further attacked by the water molecule. The study of
site-directed mutagenesis, however, indicated that the three
Asp residues, A sp353, Asp372, and Asp373, are the
potential catalytic r esidues of IPU. T he report of the crystal
structure of Dex49A also mentioned that either Asp376 or
Asp396, the residues corresponding to Asp353 and Asp373
of IPU, respectively, appears to be properly positioned to
act as a base in the hydrolytic reaction [17]. Three Asp
residues are also strictly conserved in t he catalytic centre of
GH family 28 polygalacturonases and rhamnogalacturon-
ases [18–20], another family of inverting enzymes forming
the clan GH-N w ith GH family 49 enzymes. van San ten
et al. [19] and Shimizu et al. [20] reported that a n Asp
Fig. 4. Comparison of the active site structures of the model of IPU (A) and Dex49A (B). Conserved residues between IPU and De x49A are shown in
red (mutated res idues i n this study) o r orange. Residu es that are u niqu ely found in IPU and may interact with the substrate (see Results and
Discussion), are shown in cyan. Residues that are uniquely found in Dex49A and interact with Glc +1 and +2 are shown in green. Other colour
representations a re as in Fig. 3. The figures w ere generated using
MOLSCRIPT
[40] and
RASTER
3
D
[41].
Fig. 5. Enzymatic properties of I PU. (A) Optical ro tation during the h ydrolysis of s ubstrates by a-glucosidase (top), g lu coamylase (middle), and

Glu179 and Glu400 have been reported t o function as a
general a cid a nd g eneral b ase, res pectively [4 6]. S ierks et al.
suggested that Glu179 is the general acid catalyst of pK
a
5.9,
and that the adjacent Glu180 is negatively charged, raising
the p K
a
of the general acid catalyst [23]. I t is likely t hat
catalytic residues of IPU adopt a similar catalytic mechan-
ism to those of glucoamylase.
Enzymatic properties of wild-type IPU
The modelling s tudy of the IPU structure i ndicated that the
conserved and functionally important residues of both
Dex49A and I PU are found in the negatively numbered
subsites. The anomeric configuration of products of both
Dex49A and IPU are identical, as well. Does the substrate
preference of the negatively numbered s ubsites of IPU also
resemble that of GH family 49 dextranases even though IPU
does not hydrolyse dextran at all? IPU not only h ydrolyses
panose and a polymer of panose, pullulan [1,30], but also the
oligosaccharides containing the panose structure such as
IMTG (Glc-a-(1fi6)-Glc-a-(1fi6)-Glc-a-(1fi4)-Glc) [16],
MM (Glc -a-(1fi4)-Glc-a-(1fi6) -Glc-a-(1fi4)-Glc) [1,16],
4
2
-a-isomaltosylisomaltose (Glc-a-(1fi6)-Glc-a-( 1fi4)-
Glc-a-(1fi6)-Glc) [16], and 6
3
-a-glucosylmaltotriose

panose are known as effective inducers for amylase
synthesis. In the case of A. nidulans, a mylase synthes is is
induced at an extremely low concentration ( 3 l
M
)of
isomaltose. Kato et al. reported that t wo a-glucosidases
from A. nidulans, AgdA and AgdB, showed strong trans-
glycosylation activity to produce isomaltose from maltose,
and they are suggested to participate in the maltose-
dependent induction of amylase s ynthesis along with other
undetected isomaltose-forming enzymes [49,50]. It is likely
that A. niger has a mechanism similar to such amylase
synthesis, and IPU may collaborate to produce isomaltose
from panose and other branched oligosaccharides with
some transglycosidases. Since t he extremely l ow concentra-
tion of isomaltose is effective for the induction of amylase
synthesis, the low activity of IPU may be suitable for control
of this regulation.
Acknowledgements
We thank Mr Masahiro Mizuno for his useful advice and discussions.
This study was supported in part by the Novozymes J apan R ese arch
Fund.
References
1. Sakano, Y., Higuchi, M. & Kobayashi, T. (1972) Pullulan 4-glu-
canohydrolase from Aspregillus n iger. Arch. Biochem. Biophys.
153, 1 80–187.
2. Aoki, H. & Yopi & Sakano, Y. (1997) Molecular cloning and
heterologous exp ressio n of the isopullulanase ge ne fro m Asper-
gillus n iger A.T.C.C. 964 2. Biochem. J. 323, 7 57–764.
3. Bender, H . & Wall enfels, K. (1961) Untersuchungen a n pullulan.

11. Okushima, M., Sugino, D., Kouno, Y., Nakano, S., Miyahara, J.,
Toda,H.,Kubo,S.&Matsushiro,A.(1991)Molecularcloning
and nucleotide sequencing of the Arthrobacter dextranase gene
and its expression in Escherichia coli and Streptococcus sanguis.
Jpn. J .Genet. 66, 1 73–187.
12.Roca,H.,Garcia,B.,Rodriguez,E.,Mateu,D.,Boroas,L.,
Cremata, J., Garcia, R., Pons, T. & Delgaso, D. (1996) Cloning of
the Penicillium minioluteum gene encoding dextranase and i ts
expression in Pichia pastoris. Yeast 12 , 1187–1200.
4426 H. Akeboshi et al. (Eur. J. Biochem. 271) Ó FEBS 2004
13. Garcia, B., Margolles, E., Roca, H., Mateu, D., Raices, M.,
Gonzales, M.E., He rrera, L. & Delgado, J. (1 996) Cloning and
sequencing of a dextranase-encoding cDNA from Penicillium
minioluteum. FEMS Microbiol. Lett. 143, 175–183.
14. Oguma, T., Kurokawa, T., Tobe, K., Kitao, S. & Kobayashi, M.
(1999) Clo ning and sequence analysis of the gene for gluc odex-
tranase from Arthrobacter globiformis T-3044 and expression in
Escherichia coli cells. Biosci. B iotechnol. Biochem. 63, 2174–2182.
15. Mizuno,T.,Mori,H.,Ito,H.,Matsui,H.,Kimura,A.&Chiba,S.
(1999) Molecular cloning of isomaltotrio-dextranase gene from
Brevibacterium fuscum var. dextranlyticum strain 0407 and i ts
expression in Escherichia coli. Biosci. Biotechnol. Biochem. 63,
1582–1588.
16. Akeboshi, H., Kashiwagi, Y., Aoki, H., Tonozuka, T., Nishikawa,
A. & Sakano, Y. (2003) Construction of an e fficient expression
system for Aspergillus i sopullulanase in Pichia pastoris,anda
simple purification method. Biosci. Biotec hnol. Biochem. 67, 1149–
1153.
17. Larsson, A.M., Andersson, R., S tahlberg, J., Kenne, L. & Jones,
T.A. (2003) Dextranase from Penicillium minioluteum:Reaction

T.A. (1990) Three-dimensional structure of cellobiohydrolase II
from Trichoderma r eesei. Science 249, 380–386.
25. Matsuura, Y., Kusunoki, M., Harada, W. & K akudo, M. (19 84)
Structure and possible catalytic residues of Taka-amylase A.
J. Biochem. 95, 697–702.
26. Bravman, T., Mechaly, A., Shulami, S., Belakhov, V., Baasov, T.,
Shoham, G. & Shoham, Y. (2001) Glu tamic acid 160 i s the acid-
base catalyst of b-xylosidase from Bacillus stearothermophilus T-6:
a family 3 9 glycoside hydrolase. FEB S Lett. 495, 115–119.
27. Aoki, H., Yopi, P admajanti, A. & Sakano, Y. (1996) Two com-
ponents o f cell-bound i soullulanase from Aspergil lus niger ATCC
9642-Their purification and enzymatic properties. Biosci. Bio-
technol. Biochem. 60, 1795–1798.
28. Sambrook, J., Fritsc h, E.F. & Man iatis, T . (1989) Molecular
Cloning: a Laboratory Manual, 2nd edn. Cold Spring. Harbor
Laboratory Press, Cold Spring Harbor, New York.
29. Kunkel, T.A. (19 85) Rapid and e fficient s ite-specific mutagenesis
without p henotypic selection. Proc. Natl Acad. Sci. USA 82,488.
30. Sakano, Y., Kogure, M., Kobayashi, T., Tamura, M. & Suekane,
M. (1978) Enzymatic prep aration of panose and isop an ose f ro m
pullulan. Carbohyd. Res. 61, 175–179.
31. Kim, Y.K. & Sakano, Y. (1996) Arthrobacter globiformis T6
isomaltodextranase transfers i somaltosyl res idue from dextran to
C-4 position o f acceptors. J. Appl. Glycosci. 43, 35–41.
32. Sakano, Y., Sano, M . & Ko bayashi, T. (1985) Hydrolysis o f
a-1,6-glucosid ic linkages by a-amylases. Agric. Biol. Chem. 49,
3041–3043.
33. Fox, J.D. & Robyt, J. F. (1991) Miniaturiz ation of three carbo-
hydrate analyses using a microsample plate reader. Anal. Bio chem.
195, 9 3–96.

garis and Methanococcus jannaschii whose g e nes are found in the
flanking r e gion of the a-amylase genes. Appl. Microbiol. B io tech-
nol. 56, 465–473.
44. Holm,L.&Sander,C.(1995)Dali:anetworktoolforprotein
structure comparison. Tr ends Biochem. Sci. 20, 478 –480.
45. Yokota, T., Tonozuka, T., Kamitori, S. & Sakano, Y. (2001) The
deletion of amino-terminal d omain in Thermoactinomyces vulgaris
R 47 a-amylases: Effects of domain N on activity, specificity,
stability and dimerization. Biosci. Biotechnol. Biochem. 65, 401–
408.
46. Coutinho, P.M. & Reilly , P.J. (1994) Structure-function relation-
ships in the catalytic and strach binding domains of glucoamylase.
Protein Eng. 7, 393–400.
47. Padmajanti, A., Tonozuka, T. & Sakano, Y. (2000) Deglycosy-
lated isopu llulan ase ret ains en zyma tic activ ity. J. Appl. Glycosci.
47, 2 87–292.
48. Tomomura, K., Suzuki, H., Na kamura, N ., Kuraya, K. &
Tanabe, O. (1961) On the inducers of a-amylase formation in
Aspergillus oryzae. Agric. Biol. Chem. 25, 1–6.
49. Kato, N., Murakoshi, Y., K ato, M., Kobayashi, T. & Tsukagoshi,
N. (2002) Isomaltose formed by a-glucosidase triggers amylase
induction in Aspergi llus nidulans. Curr. Genet. 42 , 43–50.
50. Kato, N., Suyama, S., Shirokane, M., Kato, M., Kobayashi, T. &
Tuskagoshi, N. (2002) Novel a-glucosidase fro m Aspe rgillus
nidulans with strong transglycosylation activity. Appl. E nviron.
Microbiol. 68, 1250–1256.
Ó FEBS 2004 Reaction mechanism of isopullulanase (Eur. J. Biochem. 271) 4427