Probing the active site of Corynebacterium callunae starch
phosphorylase through the characterization of wild-type
and His334
fi
Gly mutant enzymes
Alexandra Schwarz
1
, Lothar Brecker
2
and Bernd Nidetzky
1
1 Institute of Biotechnology and Biochemical Engineering, Graz University of Technology, Austria
2 Institute of Organic Chemistry, University of Vienna, Austria
Glycogen phosphorylases are pyridoxal 5¢-phosphate
(PLP)-dependent glycosyltransferases (EC 2.4.1.1) that
catalyze the reversible phosphorolysis of oligomeric
and polymeric a-1,4-glucan substrates (maltodextrins,
starch, glycogen) [1,2]. The reaction proceeds with
retention of configuration at the anomeric carbon,
yielding a-d-glucose 1-phosphate (Glc1P) as product
in the direction of substrate depolymerization. In spite
Keywords
a-retaining glucosyl transfer; phosphorus
NMR; pyridoxal 5¢-phosphate; saturation
transfer difference NMR; starch
phosphorylase
Correspondence
B. Nidetzky, Institute of Biotechnology and
Biochemical Engineering, Graz University of
Technology, Petersgasse 12, A-8010 Graz,
Austria

tion transfer difference NMR experiments, suggested that disruption of
enzyme–substrate interactions in H334G was strictly local, affecting the
protein environment of sugar carbon 6. pH profiles of the phosphorolysis
rate for wild-type and H334G were both bell-shaped, with the broad pH
range of optimum activity in the wild-type (pH 6.5–7.5) being narrowed
and markedly shifted to lower pH values in the mutant (pH 6.5–7.0).
External imidazole partly restored the activity lost in the mutant, without,
however, participating as an alternative nucleophile in the reaction. It
caused displacement of the entire pH profile of H334G by + 0.5 pH units.
A possible role for His334 in the formation of the oxocarbenium ion-like
transition state is suggested, where the hydrogen bond between its side
chain and the 6-hydroxyl polarizes and positions O-6 such that electron
density in the reactive center is enhanced.
Abbreviations
CcStP, Corynebacterium callunae starch phosphorylase; GL,
D-gluconic acid 1,5-lactone; Glc1P, a-D-glucose 1-phosphate; LFER, linear free
energy relationship; PLP, pyridoxal 5¢-phosphate; STD, saturation transfer difference; X1P, a-
D-xylose 1-phosphate.
FEBS Journal 274 (2007) 5105–5115 ª 2007 The Authors Journal compilation ª 2007 FEBS 5105
of detailed studies spanning many decades, definite
conclusions about the catalytic mechanism of glycogen
phosphorylases and the exact function of the PLP co-
factor in it are still elusive [2–6]. Figure 1 shows that
an active site His has a central role in the contentious
debate surrounding a putative covalent glucosyl–
enzyme intermediate of a double displacement-like
mechanism of the phosphorylase. The precedent of
sucrose phosphorylase [7–9], mechanistically represent-
ing a large class of retaining glycoside hydrolases and
transglycosidases, would strongly favor some form of

syl oxoarbenium ion mimics such as d-gluconic acid
1,5-lactone (GL) [16,17]. Substitution of His334 in
starch phosphorylase from Corynebacterium callunae
(CcStP) (Fig. 1A) by Gln or Asn caused a substantial
(up to 150-fold) loss in wild-type catalytic efficiency
that was paralleled by a corresponding decrease in
affinity for GL in combination with phosphate, reflect-
ing a change from positive to negative cooperativity in
binding of the two ligands as a result of the site-direc-
ted replacement [18].
In this work, we have substituted His334 with Gly
and analyzed the disruptive effects of the point muta-
tion on active site function of CcStP using steady-state
kinetics and selective NMR probes for the 5¢-phos-
phate group of the cofactor and for bound carbo-
hydrate ligands. The work was carried out to address
three questions in particular, taking into account that,
quite unexpectedly, an H334A mutant of CcStP was
almost as active as the wild-type enzyme [18]. How
does complete removal of the side chain of His334
influence binding and catalysis? Are the properties of
neighboring active site groups, including the PLP
cofactor, affected by the His fi Gly mutation? If suffi-
cient room is vacated in H334G to accommodate
water or another nucleophile in place of the original
methylimidazole group, will this new ligand participate
in the enzymatic reaction such that eventually hydro-
N
O
N

O
O
N
O
N
O
O
His345 (334)
Gly114 (114)
Leu115 (115)
Gly640 (629)
Glu637 (626)
Tyr538 (527)
Asn449 (437)
Ser639 (628)
3.0 Å
3.1 Å
2.9 Å
2.9 Å
3.6 Å
3.6 Å
3.6 Å
2.7 Å
3.1 Å
3.7Å
2.7Å
2.8 Å
3.2 Å
4.3 Å
3.2 Å

(33 UÆmg
)1
). External imidazole stimulated activity of
the mutant up to 5.5-fold, whereas it weakly inhibited
the wild-type (Fig. 2). Acetate and formate had no
effect on the activity of the H334G mutant. Azide,
2-methylimidazole and 2-ethylimidazole inhibited the
mutant. The wild-type was inhibited weakly (£ 2-fold)
by all of the compounds tested, with the exception of
formate, which caused a five-fold reduction of activity.
Kinetic parameters
Steady-state kinetic parameters for phosphorolysis of
maltopentaose by the H334G mutant were determined
at pH 7.0 under conditions where the concentration of
phosphate was constant and saturating (50 mm). The
k
cat
of 0.033 ± 0.001 s
)1
was 0.05% of the wild-type
value. The K
m
for maltopentaose was 280 ± 20 mm,
reflecting a 75-fold decrease in substrate binding affin-
ity as a result of the mutation. Like the wild-type [18],
the H334G mutant did not hydrolyze maltopentaose
into glucose above a detection limit of about 0.15% of
its phosphorylase activity.
Ligand binding
Dissociation constants (K

trast, the pH rate profile of the wild-type was not
affected by addition of the same concentration of imid-
azole (data not shown).
Phosphorus NMR of pyridoxal 5¢-phosphate
31
P-NMR spectra for solutions of wild-type CcStP and
the H334G mutant that contained a similar concentra-
tion of enzyme-bound PLP ( 100 lm) were recorded
in the pH range 5.6–8.0. Typical spectra acquired at
pH 7.25 are shown in Fig. 4A. The
31
P resonance
imidazole (mM)
0
rel. activity (-fold)
0
2
4
6
100 200
300
400 500
Fig. 2. Analysis of restoration of activity in wild-type CcStP (d) and
the H334G mutant (s) by external imidazole. The results are given
as relative specific activities that were normalized by using the spe-
cific activities of the wild-type (33 UÆmg
)1
) and the H334G mutant
(0.001 UÆmg
)1

enzyme were remarkably insensitive to the binding of
arsenate alone and in combination with GL.
Analysis of ligand binding by STD NMR
Figure 5 summarizes relative saturation transfer differ-
ence (STD) effects of Glc1P and a-d-xylose 1-phos-
phate (X1P) upon their binding to wild-type and
H334G phosphorylase. Glc1P displayed very similar
patterns of binding to both enzymes. However, the
relative STD effects of the protons in positions 6a
and b were slightly higher when Glc1P was bound to
the wild-type than when it was bound to the H334G
mutant. The relative STD effects of X1P bound to
the two enzymes were also fairly similar, with the
exception of the proton in position 5eq, which showed
a higher effect in the complex with the H334G
mutant. Binding of GL to the wild-type and the
H334G mutant also yielded very similar STD spectra
with, however, quite a low signal-to-noise ratio, very
likely caused by the small dissociation constants for
enzyme–GL complexes. Appreciable STD effects could
be detected only for protons in positions 2 and 4,
which caused overlapping signals in the
1
H-NMR
spectrum (data not shown) [20]. All other protons
showed much lower STD effects, which could not be
quantified. Although longer STD measurements could,
in principle, improve the signal-to-noise ratio, the
duration of the NMR experiment was limited in this
case by the spontaneous hydrolysis of GL to gluconic

presence of 20 m
M arsenate and 1 mM GL
(.);
31
P shift for the H334G mutant (,),
recorded in the absence of arsenate and at
only a single pH of 7.25.
p
H
5.5
log(rel. k
cat
)
1.4
1.6
1.8
2.0
6.0 6.5 7.0 7.5 8.0
Fig. 3. pH profiles of catalytic rates for phosphorolysis of maltodex-
trin catalyzed by wild-type CcStP (.) and the H334G mutant in the
absence (d) and presence (s) of 200 m
M imidazole. The initial
rates were acquired under conditions of apparent saturation with
substrate, and are given as relative values (rel. k
cat
) of the catalytic
rate for the wild-type (50 s
)1
; pH 7.0) and the catalytic rates of the
H334G mutant in the absence (0.0015 s

electrostatic interactions are silent in the STD NMR
experiment, the obtained portrait of the binding pat-
tern is partial (Fig. 1B), and isolated interpretations of
STD effects can therefore be hazardous. However, if
STD effects for two minimally modified systems can
be investigated and compared, then the interpretation
is considerably simplified. The side chain of His334
and the –CH
2
OH group of Glc1P are complementary
interacting groups (Fig. 1B), and analysis of changes
in relative STD effects resulting from structural pertur-
bation of enzyme (H334G) and substrate (X1P) was
therefore of particular interest. The results obtained
suggest an overwhelmingly local disruption of binding
interactions caused by removing the two functional
groups individually or together.
Analysis of kinetic consequences in the H334G
mutant and chemical rescue studies
Substitution of His334 with Gly caused a 10
3.5
-fold
decrease in the wild-type k
cat
for phosphorolysis of
maltopentaose. Conversion of the ternary enzyme–sub-
strate complex is believed to be the rate-determining
step of glucosyl transfer to phosphate catalyzed by
a-glucan phosphorylases [1], and k
catP

reference proton spectrum. The effects are
normalized to the respective largest effect
in the sample.
A. Schwarz et al. Role of His334 in a-glucan phosphorylase
FEBS Journal 274 (2007) 5105–5115 ª 2007 The Authors Journal compilation ª 2007 FEBS 5109
calculated with the relationship DDG# ¼ RT ln 10
5.2
,
using the ratio of k
catP
⁄ K
m
values of 18 000 m
)1
Æs
)1
and 0.12 m
)1
Æs
)1
for the wild-type and the H334G
mutant, respectively.) We speculated that water might
occupy the position vacated in the H334A mutant
through removal of the imidazole group of the His,
thereby effectively replacing the function of the origi-
nal side chain in catalysis by the mutant [18]. What-
ever mechanism truly accounts for the retention of
phosphorylase activity by the H334A mutant, it is
clearly not available to the H334G mutant. The selec-
tivity of the H334G mutant for glucosyl transfer to

to cause a breakdown of the LFER, in contrast to the
observations made. Whereas external imidazole weakly
enhanced the activity of the H334G mutant, it did not
participate in the reaction as alternative nucleophile,
such that glucose 1-imidazole or the product of its
spontaneous hydrolysis (glucose) would be formed in
kinetic competition with Glc1P. Other small nucleo-
philes, such as azide, were without effect on both
activity and reaction course. By way of comparison,
when the catalytic nucleophile (Asp) of sucrose phos-
phorylase was replaced by Ala, azide could occupy the
position of the original carboxylate group and react
through addition to C-1 of the glucosyl moiety, yield-
ing the inversion product b-glucose 1-azide [9].
We investigated whether the proposed hydrogen
bond between His334 and the C-6 hydroxy group of
the glucosyl residue bound at the catalytic subsite
could become optimized in the transition state. A
hypothetical scenario, inspired by studies of human
purine nucleoside phosphorylase [29,30], is that His334
could be responsible for positioning O-6 in line with
O-5 and the glycosidic oxygen of phosphate (O
P1
)
(Fig. 6). In the direction of polysaccharide synthesis,
compression of the three-oxygen stack such that O-6
moves closer to the ring oxygen would enhance elec-
tron density in the reactive carbon and thus facilitate
glycosidic bond cleavage and formation of the transi-
tion state in an S

subsite in the selective stabilization of the transition state. O-6,
the ring oxygen, and the glycosidic oxygen O
P1
lie in a close three-
oxygen stack that is indicated by a dashed line. Increased electron
density near the reactive center provided by squeezing the three
oxygens together could facilitate the catalytic step. The picture was
generated using Protein Data Bank entry 1L5V (maltodextrin phos-
phorylase bound with Glc1P [15]).
Role of His334 in a-glucan phosphorylase A. Schwarz et al.
5110 FEBS Journal 274 (2007) 5105–5115 ª 2007 The Authors Journal compilation ª 2007 FEBS
bell-shaped curves showing a decrease in activity at
low and high pH. Replacement of His334 with Gly
caused a marked change in the pH profile of k
cat
for
the phosphorolysis direction. To explore possible
sources of the different pH dependences, we used
31
P-NMR and compared chemical shifts for the
5¢-phosphate group of PLP in the wild-type and the
H334G mutant. Changes in chemical shift and line
width of the
31
P-NMR signal may serve as reporters
of alterations in the ionization state of the cofactor
phosphate group [36]. They are, however, also expli-
cable by changes in the local environment of PLP
and their effect on conformational strain on the
5¢-phosphate moiety.

phosphate remains essentially unaffected upon forma-
tion of enzyme complexes with arsenate alone and in
combination with GL. By contrast, significant field
shifts of the
31
P resonance signal were observed with
E. coli maltodextrin phosphorylase [37], potato phos-
phorylase [39] and muscle glycogen phosphorylase
[40] upon addition of arsenate, probably caused by
electrostatic interactions between the 5¢-phosphate
moiety and arsenate. The pK
a
for PLP phosphate in
E. coli maltodextrin phosphorylase was also shifted
by + 1.1 pH units upon binding of arsenate [37].
Therefore, CcStP appears to differ subtly from mal-
todextrin and glycogen phosphorylase in how it copes
with constraining the cofactor phosphate group into a
configuration that is believed to promote catalysis via
direct interaction with the substrate arsenate (or
phosphate). A tentative explanation is provided by
Fig. 7, which reveals clear differences in the pattern
of hydrogen bonding and the orientation of PLP
phosphate in the active sites of CcStP bound with
phosphate (Fig. 7A) and maltodextrin phosphorylase
bound with phosphate and a nonphosphorolyzable
substrate analog (omitted in Fig. 7B for reasons of
clarity) in Fig. 7B. Gly642 in the E. coli enzyme is
substituted by Ser631 in CcStP. Interactions from the
main chain amide of Gly are replaced by interactions

substrate analog [15]).
A. Schwarz et al. Role of His334 in a-glucan phosphorylase
FEBS Journal 274 (2007) 5105–5115 ª 2007 The Authors Journal compilation ª 2007 FEBS 5111
pH 6.5 [19] may be correlated, at least formally, with
the strong field shift of
31
P resonance signal in this pH
range, perhaps reflecting the formation of a PLP dian-
ion. Electrostatic repulsion may now prevent the cofac-
tor 5¢-phosphate and also the dianionic phosphate of
the glucosyl donor substrate from closely approaching
each other [2,41].
Rather than eliminating a single ionization from
pH profiles, substitution of His334 by Gly caused a
complex pattern of changes in the pH rate dependenc-
es of the wild-type. The acidic and basic limbs on the
pH profile of the H334G mutant for the phosphoroly-
sis direction were displaced inward by  0.5 pH units
in comparison with the corresponding pH profile of
the wild-type, and the optimum pH range for the
mutant was also shifted, by about ) 0.75 pH units. In
addition to partly restoring activity in the H334G
mutant, external imidazole caused an upshift by £ 1.0
pH units of the entire pH dependence of phosphoro-
lysis by the mutant, whereas the pH rate profile of
the wild-type was not influenced by the added imidaz-
ole. Although these results suggest that His334 influ-
ences the pH dependence of the activity of CcStP,
they do not delineate a detailed relationship. Appar-
ent ionizations on the pH rate profiles must probably

ried out using published protocols [18]. Enzyme activity
was measured with a continuous coupled assay reported
elsewhere [9], and protein was determined by the Bio-Rad
(Vienna, Austria) dye binding assay using BSA as standard.
Steady-state kinetic analysis and biochemical
characterization
Initial rates of phosphorolysis were determined in discontin-
uous assays as described previously [44]. The enzyme (5.5 lm
subunits of the H334G mutant) was incubated at 30 °Cin
300 mm potassium phosphate buffer, and the release of
Glc1P was measured as a function of time of incubation up
to 3 h. Maltodextrin or maltopentaose was used as the sub-
strate, as indicated in Results. The sodium salts of azide, ace-
tate, and formate, as well as imidazole, 2-ethylimidazole, and
2-methylimidazole, were tested in the range 10–250 mm for
possible restoration of activity of the H334G mutant for
phosphorolysis of maltodextrin (23 gÆL
)1
) at pH 7.0. Con-
trol reactions with the wild-type were carried out in all cases.
The H334G mutant was examined for possible hydrolase
activity by incubating the enzyme (6.7 lm)at30°Cin
50 mm triethanolamine buffer (pH 7.0), containing malto-
pentaose (75 mm) and potassium sulfate (20 mm). Note that
sulfate was added in this series of measurements to ensure
stability of the enzymes during the timespan of experiments
carried out in the absence of phosphate [45]. Samples were
taken at certain time points up to 40 h, and the formation of
glucose was measured as described elsewhere [18].
pH dependence studies were performed in the pH range

H) at 30 °C on a Bruker (Rheinstetten, Germany)
DRX 600 AVANCE spectrometer using topspin 1.3 soft-
ware (Bruker). Proton, carbon and phosphorus spectra
were measured at 600.13 MHz, 150.90 MHz, and
242.94 MHz, respectively. The one-dimensional spectra
were recorded with 32 768 data points. Zero filling to
65 536 data points, appropriate exponential multiplication
and Fourier transformation led to spectra with ranges of
 5400 Hz (
1
H),  33 000 Hz (
13
C), and  24 000 Hz (
31
P).
pH-dependent
31
P chemical shifts were determined employ-
ing a slight modification of a reported procedure [37].
Samples were prepared by adding 150 lL of 400–450 lm
enzyme solution in 50 mm triethanolamine buffer, contain-
ing 20 mm potassium sulfate, to 450 l L of a solution con-
taining 50 mm triethanolamine buffer, 50 mm acetate and
20 mm potassium sulfate in
2
H
2
O. At the concentration
used, sulfate does not inhibit the enzyme activity through
competition with phosphate, suggesting that occupancy of

O (pH 6.65). Five
hundred and twelve scans were collected, each with 50
Gaussian-shaped pulses (50 ms and 1 ms delay) and a
30 ms spin lock pulse, resulting in spectra of  4200 Hz
spectral width. On and off resonance irradiations were
made at d
H
) 2.00 p.p.m. and d
H
41.66 p.p.m., respectively,
subtraction was performed via phase cycling, and no water
suppression was applied. Reference proton spectra were
recorded with 256 scans directly before and after the STD
measurements.
Acknowledgements
Financial support from the Austrian Science Fund
(FWF P15208-B09, P18038-B09 and P15118) is grate-
fully acknowledged.
References
1 Graves DJ & Wang JH (1972) a-Glucan phosphorylases
) chemical and physical basis of catalysis and regula-
tion. In The Enzymes (Boyer PD, ed.), pp. 435–483.
Academic Press, New York, NY.
2 Palm D, Klein HW, Schinzel R, Buehner M & Helmr-
eich EJ (1990) The role of pyridoxal 5¢-phosphate in
glycogen phosphorylase catalysis. Biochemistry 29,
1099–1107.
3 Madsen NB & Withers SG (1986) Pyridoxal phosphate
and derivatives. In Coenzymes and Cofactors (Dolphin
D, Paulson R & Avramovic O, eds), pp. 355–389.

Garman EF, Oikonomakos NG & Johnson LN (1996)
Ternary complex crystal structures of glycogen
phosphorylase with the transition state analogue
nojirimycin tetrazole and phosphate in the T and R
states. Biochemistry 35, 7341–7355.
12 O’Reilly M, Watson KA, Schinzel R, Palm D & John-
son LN (1997) Oligosaccharide substrate binding in
Escherichia coli maltodextrin phosphorylase. Nat Struct
Biol 4, 405–412.
13 Watson KA, McCleverty C, Geremia S, Cottaz S, Dri-
guez H & Johnson LN (1999) Phosphorylase recogni-
tion and phosphorolysis of its oligosaccharide substrate:
A. Schwarz et al. Role of His334 in a-glucan phosphorylase
FEBS Journal 274 (2007) 5105–5115 ª 2007 The Authors Journal compilation ª 2007 FEBS 5113
answers to a long outstanding question. EMBO J 18,
4619–4632.
14 O’Reilly M, Watson KA & Johnson LN (1999) The
crystal structure of the Escherichia coli maltodextrin
phosphorylase–acarbose complex. Biochemistry 38,
5337–5345.
15 Geremia S, Campagnolo M, Schinzel R & Johnson LN
(2002) Enzymatic catalysis in crystals of Escherichia coli
maltodextrin phosphorylase. J Mol Biol 322, 413–423.
16 Gold AM, Legrand E & Sanchez GR (1971) Inhibition
of muscle phosphorylase a by 5-gluconolactone. J Biol
Chem 246, 5700–5706.
17 Tu JI, Jacobson GR & Graves DJ (1971) Isotopic
effects and inhibition of polysaccharide phosphorylase
by 1,5-gluconolactone. Relationship to the catalytic
mechanism. Biochemistry 10, 1229–1236.

25 Yao H & Sem DS (2005) Cofactor fingerprinting with
STD NMR to characterize proteins of unknown func-
tion: identification of a rare cCMP cofactor preference.
FEBS Lett 579, 661–666.
26 Haselhorst T, Mu
¨
nster-Ku
¨
hnel AK, Stolz A, Oschlies
M, Tiralongo J, Kitajima K, Gerardy-Schahn R & von
Itzstein M (2005) Probing a CMP-Kdn synthetase by
1
H,
31
P, and STD NMR spectroscopy. Biochem Biophys
Res Commun 327, 565–570.
27 Brecker L, Straganz GD, Tyl CE, Steiner W & Nidetzky
B (2006) Saturation-transfer-difference NMR to charac-
terize substrate binding recognition and catalysis of two
broadly specific glycoside hydrolases. J Mol Catal B:
Enzym 42, 85–89.
28 Segel IH (1993) Rapid equilibrium bireactant and terre-
actant systems. In Enzyme Kinetics , pp. 273–345. John
Wiley & Sons, New York, NY.
29 Nunez S, Antoniou D, Schramm VL & Schwartz SD
(2004) Promoting vibrations in human purine nucleoside
phosphorylase. A molecular dynamics and hybrid quan-
tum mechanical ⁄ molecular mechanical study. JAm
Chem Soc 126, 15720–15729.
30 Murkin AS, Birck MR, Rinaldo-Matthis A, Shi W &

P reporter observing functional
changes in the active site of Escherichia coli maltodex-
trin phosphorylase after site-directed mutagenesis.
Biochemistry 31, 4128–4133.
38 Becker S, Schnackerz KD & Schinzel R (1995) A study
of binary complexes of Escherichia coli maltodextrin
phosphorylase: a-d-glucose 1-methylenephosphonate as
a probe of pyridoxal 5¢-phosphate–substrate interac-
tions. Biochim Biophys Acta 1243, 381–385.
39 Klein HW & Helmreich EJ (1979) A proton donor–
acceptor function of the 5¢-phosphate group of pyri-
doxal-P in potato phosphorylase inferred from
31
P
NMR spectra. FEBS Lett 108, 209–214.
40 Feldmann K & Hull WE (1977)
31
P nuclear magnetic
resonance studies of glycogen phosphorylase from
Role of His334 in a-glucan phosphorylase A. Schwarz et al.
5114 FEBS Journal 274 (2007) 5105–5115 ª 2007 The Authors Journal compilation ª 2007 FEBS
rabbit skeletal muscle: ionization states of pyridoxal
5¢-phosphate. Proc Natl Acad Sci USA 74, 856–860.
41 Withers SG, Madsen NB & Sykes BD (1981) Active
form of pyridoxal phosphate in glycogen phosphorylase.
Phosphorus-31 nuclear magentic resonance investiga-
tion. Biochemistry 20, 1748–1756.
42 Griessler R, Schwarz A, Mucha J & Nidetzky B (2003)
Tracking interactions that stabilize the dimer structure
of starch phosphorylase from Corynebacterium callunae.

phate complexes of wild-type CcStP and His334
mutants and k
cat
⁄ K
m
values for phosphorolysis of
maltopentaose (A) and starch (B).
This material is available as part of the online article
from http://www.blackwell-synergy.com
Please note: Blackwell Publishing is not responsible
for the content or functionality of any supplementary
materials supplied by the authors. Any queries (other
than missing material) should be directed to the corre-
sponding author for the article.
A. Schwarz et al. Role of His334 in a-glucan phosphorylase
FEBS Journal 274 (2007) 5105–5115 ª 2007 The Authors Journal compilation ª 2007 FEBS 5115