Prorein Science (1997), 6:294-303. Cambridge University Press. Printed in the USA.
Copyright 0 1997 The Protein Society
Three-dimensional structures of three engineered
cellulose-binding domains of cellobiohydrolase I
from Trichoderma reesei
MAIJA-LIISA MATTINEN,' MAARIT KONTTELI,' JANNE KEROVUO,' MARKUS LINDER?
ART0 ANNILA,' GUNNAR LINDEBERG? TAPANI REINIKAINEN;
AND TORBJORN DRAKENBERG'
'VTT, Chemical Technology, Box 1401, FIN-02044 VTT, Finland
*VTT, Biotechnology and Food Research, Box 1500, FIN-02044 VTT, Finland
'Department of Medical and Physiological Chemistry, University of Uppsala, Box 575, S-75124 Uppsala, Sweden
July 31, 1996: ACCEFTEO November 20, 1996)
Three-dimensional solution structures for three engineered, synthetic CBDs (Y5A, Y31 A, and Y32A) of cellobiohydro-
lase I (CBHI) from Trichoderma reesei were studied with nuclear magnetic resonance (NMR) and circular dichroism
(CD) spectroscopy. According to CD measurements the antiparallel P-sheet structure of the CBD fold was preserved in
all engineered peptides. The three-dimensional NMR-based structures of Y31A and Y32A revealed only small local
changes due to mutations in the flat face of CBD, which is expected to bind to crystalline cellulose. Therefore, the
structural roles of Y3 1 and Y32 are minor, but their functional importance is obvious because these mutants do not bind
strongly to cellulose. In the case of Y5A, the disruption of the structural framework at the N-terminus and the complete
loss of binding affinity implies that Y5 has both structural and functional significance. The number of aromatic residues
and their precise spatial arrangement in the flat face of the type I CBD fold appears to be critical for specific binding.
A model for the CBD binding in which the three aligned aromatic rings stack onto every other glucose ring of the
cellulose polymer is discussed.
Keywords: cellulase; nuclear magnetic resonance spectroscopy; protein-carbohydrate interaction; structure
In nature, cellulose is degraded by a variety of micro-organisms.
The filamentous fungus Trichoderma reesei has been widely used
as a model organism in studies of cellulose breakdown (Quiocho,
1986, 1993; Claeyssens & Tomme, 1989; Vyas, 1991; Poole et al.,
1993; Din et al., 1994). T. reesei secretes a mixture of cellulases
consisting of two cellobiohydrolases (CBHI and CBHII) and at
least three endoglucanases (EGI, EGII, and EGV), which have
been shown to act synergistically (Henrissat et al., 1985; Wood &
Box 1400, FIN-02044 VTT, Finland; e-mail: firstname.lastname@example.org.
Abbreviations: 2D, two-dimensional: CBD, cellulose-binding domain;
Reprint requests to: Maija-Liisa Mattinen, VTT, Chemical Technology,
type I CBD of EGI: CBH, cellobiohydrolase; CBHI, cellobiohydrolase I:
CBHII, cellobiohydrolase 11; CD, circular dichroism: Cex, exoglucanase/
xylanase from Cellulornonas fimi; COSY, correlation spectroscopy: DG,
distance geometry: EG, endoglucanase; EGI, endoglucanase I: EGII, en-
doglucanase 11; EGV, endoglucanase V; FID, free induction decay: MD,
molecular dynamics: NMR, nuclear magnetic resonance: NOE, nuclear
Overhauser enhancement: NOESY, nuclear Overhauser enhancement spec-
troscopy: RELAY-COSY, relayed coherence transfer correlation spectros-
copy: SA, simulated annealing: TOCSY, total correlation spectroscopy.
type I CBD of CBHI; CBDc,,, type I1 CBD of Cex; CBDEGI,
Garcia-Campayo, 1990; Irwin et al., 1993; Nidetzky et al., 1993;
BCguin & Aubert, 1994). These enzymes have a catalytic core
domain and a type I cellulose-binding domain (CBD), which are
connected by a glycosylated linker peptide.
Much of the current knowledge on cellulase-cellulose binding
stems from separate studies of the soluble enzyme and the solid
substrate because the complete biphasic system is difficult to in-
The core domains of both CBHI and CBHII are thought to
degrade solubilized chains by attacking the free ends and sub-
sequently releasing cellobiosyl units. EGs hydrolyse amorphous
microcrystalline cellulose by cleaving the polymer chains at ran-
dom. P-glucosidases complete the hydrolytic process yielding glu-
cose to be metabolized by the organism (Cambillau & van Tilbeurgh,
The catalytic core has some binding affinity for crystalline cel-
lulose in the absence of even CBD, but it is much lower than that
of the intact enzyme, whereas detached CBD alone binds strongly
to cellulose. Thus, CBD is required for the full activity of cellu-
lases on crystalline cellulose (van Tilbeurgh et al., 1986; Stihlberg
et al., 1988, 1991; Tomme et a]., 1988; Reinikainen et al., 1992,
Structureactivity relationship o f engineered CBDcBHl
1995), however, the detailed binding mechanism of the CBD is not
The three-dimensional structure of CBDcaHl has been deter-
mined by NMR spectroscopy (Kraulis et al., 1989). The main
secondary structure of CBDceHl is an irregular anti-parallel triple-
stranded &sheet. The peptide folds into a wedge-shaped structure.
One face of the wedge is flat and hydrophilic, while the other face
is rougher and less hydrophilic. Although both faces contain seg-
ments of conserved amino acids, it is the flat face with three
aromatic residues that is presumed to bind to cellulose (Linder
et al., 1995a, 1995b; Reinikainen et al., 1995). This conclusion has
been supported by point mutation experiments. When the tyrosines
on the flat face were replaced one by one by alanine, the binding
affinity to cellulose was dramatically reduced (Linder et al., 1995a).
The two amide groups neighboring the tyrosines were also shown
to contribute to the binding strength (Linder et al., 1995a); how-
ever, the structural consequences induced by the mutations re-
In the present work. the structural changes caused by the point
mutations (Y5A, Y31A, and Y32A) were examined by determin-
ing the three-dimensional structures of the peptides.
Qualitatively the CD spectra of the mutants resemble that of the
wild type, showing that the secondary structures of the engineered
peptides are similar to those of the wild type (Fig. I). All spectra
have a band in the region around 205-250 nm, which is charac-
teristic of an anti-parallel P-sheet. Otherwise, the spectra are not
typical for a P-sheet structure due to short strands and many type
1 turns in the CBD fold. The low amplitude and merely qualitative
spectral characteristics do not allow any detailed conclusions to be
drawn regarding structural changes in the engineered peptides as
compared to the wild type.
As has been shown, earlier chemical shifts of the mutants differed
little from shifts of the wild type (Linder et al., 1995a).
Fig. 1. Near-UV CD spectra of Y5A (diamond), Y31A (square), Y32A
(triangle), and wild type (-).
The spectrum of the background (solvent)
has been subtracted from the spectra of the samples.
This is, however, not necessarily an indication of small structural
changes because in a small domain like CBD a majority of 'H
shifts do not deviate strongly from the random coil shifts. In gen-
eral, the NMR spectra of Y3 IA and Y32A are very similar to those
of the wild type. For Y5A, certain resonances are clustered and the
number of cross-peaks in the NOESY spectra is significantly smaller
From the final sets of calculated structures, only those with at most
two violations of no more than 0.4 A were accepted for subsequent
examination (Fig. 3). For Y5A, 37 structures were accepted. The
average of the calculated energies for these structures is about 70
kcal/mol. For Y3 1 A and Y32A, using the same selection criteria as
for Y5A, 18 and 19 structures were accepted, respectively. The
average of the calculated energies for these structures is about 20
kcal/mol. Qualitatively for the less-ordered structure of Y5A the
calculated energy is larger than for the well-structured Y31A and
Y32A. The calculated energy for the wild type is even smaller. The
number of violated distance constraints is comparable for all of the
engineered peptides (Table 1).
A comparison of the engineered CBDs with the wild type shows
an unmistakable resemblance. The resolution of Y3 IA and Y32A
allows a close examination. For Y31A and Y32A the rms devia-
tions of the backbone structures (0.53 A for Y31A and 0.81 A for
Y32A) are only slightly larger than that of the wild type (0.33 A)
(Table I). The side chains are determined to a resolution of about
1 A. The structures of Y5A have on an average a clearly inferior
resolution (-2 A) for the backbone atoms and also larger disper-
sion for the side chains (-2.6 A). For all the engineered peptides,
the rms deviation per residue correlate well with the number of
NOES per residue (Fig. 4). For Y3 1A and Y32A, the rms deviation
is small and fluctuates by about 1 A along the whole sequence,
whereas for Y5A it vary much more along the sequence. The
deviation is largest in the N-terminus and between residues (315-
C25. Even the relative position of the N-terminus with respect to
loop S 14-C 19 remains uncertain. Consequently, the structure gen-
eration resulted in two sets of families one for each topological
possibility. N-terminus is either above or below the plane of the
turn. The segment from 4 7 to G15 and 426 to L36, on the other
hand, have comparatively small (-2 A) rms deviations. For all the
engineered peptides the @sheets are better defined than the loops
The structural consequences of mutations to the flat face are seen
in Figures 5 and 6 where the backbones of the engineered peptides
are superimposed on the wild type. Only the backbone trace of the
structure with the smallest rrns deviation with respect to the wild
type is shown for clarity. In Figure 5 the side chains of the residues
on the flat face (5, 3 1, 32,29,34) are drawn for all members of the
families to reveal the resolution. The resolution of the wild-type
structure is comparable; however, the side chains are shown for
only one of these structures.
The flat face of Y31A is quite well preserved. A3 1 has moved
only slightly towards N29 as compared to Y3 I, and there are only
minor rearrangements in the remaining residues of the flat face
compared to the wild type. The aromatic rings of Y5 and Y32
occupy slightly different positions compared to the wild type. Y5
M.-L.. Mattinen et al.
0 e a -
.-....I... . . . ..I. . e . . . . W." .
9.. .. a .
- -.. ..*' ..
Fig. 2. NH (61 axis)-aliphatic (82 axis) regions of the 2D 1H-1H NOESY spectra of the engineered peptides (red Y5A, blue: Y31A,
and black: Y32A). The spectra are superimposed to facilitate comparison. Note that the wild type is very similar to Y31A and therefore
is tilted about 30 degrees from the plane of the flat face. N29 and
Q34 are in their wild-type-like positions within the precision
In the case of Y32A, the flat face has become concave and A32
is buried in the interior. Due to the mutation, the residues that
neighbor Y32 in the wild-type structure have obtained more space.
The amide groups of 434 and N29 have moved
other filling out the space left by Y32. Also, Y31 and Y5 have
moved towards each other by about 5 A and their rings have tilted
out of the flat face of the wild type. Thus, this engineered CBD has
neither the planarity nor the periodicity of the residues in the flat
face of the wild type.
In the case of Y5A the flat face is less precisely defined com-
pared to the other engineered CBDs, because there
experimental data to determine the position
relative to the plane of the loop S14-Cl9. When the N-terminus is
below the plane, A5 is approximately in the plane of the flat face,
whereas in the other case it is far above. For both families
structures the remaining residues of the flat face have average
positions similar to those in the wild type.
closer to each
is too little
of the N-terminus
The CBDCsHl binds on
dependence is weak (Reinikainen et al. 1995). The compact triple-
slightly acidic conditions, and the pH
stranded anti-parallel P-sheet of the type I CBD fold constitutes a
framework for the functionally important residues
The backbones of the engineered CBDs Y31A and Y32A follow
the fold of the wild type (Fig. 3). The strands and turns
preserved. Furthermore, the backbones are well defined along the
entire sequence (Fig. 4 ) , and most of the side-chain positions are
well defined for these peptides. This implies that the mutations of
Y31 and Y32 neither alter nor decrease the stability of the CBD
fold, i.e., they are not part of the structural framework.
In contrast, the mutation of Y5 to alanine has obvious structural
consequences. Although segments that correspond to the anti-
parallel P-strands and the connecting turns remain rather well de-
fined, the N-terminus as well as the spatially close segment G15-
C25 have poor definitions. The structure definition is for this part
of the peptide so poor that the mutual positions of the N-terminus
and the P2-strand (residues 24-28) remain ambiguous. For Y5A,
approximately half of the accepted structures have the N-terminus
above the loop 14-19 as the wild type, whereas the other half have
it below. In general, Y5A has less order than
may imply that the overall stability of the structure is compro-
mised. It is concluded that Y5 plays a key role in maintaining the
To ensure reliability of the Y5A results, 17 additional DGII
structures were calculated without any &angle restraint for A5. It
was suspected that this restraint determines the location
N-terminus due to the probable structural role of tyrosine 5. Nev-
on the flat face.
Y3 1A and Y32A. This
Structure-activity relationship of engineered CBDcsHl
Fig. 3. Structures of Y5A, Y31A, and Y32A. Only backbone C"-traces are shown for clarity. For Y5A the backbone (N, C", C, 0)
atoms of residues 7-15 and 25-36 are superimposed. (A) 23 and (B) 14 structures belonged to two different structural sets. For Y31A
(C) 18 and for Y32A (D) 19 correspondingly superimposed structures are shown.
ertheless, these calculations reproduced the two structural sets for
the N-terminus. Twelve structures belonged to the wild-type-like
family and the remaining five to the other. A calculation with the
4-angle restraint of the wild type also resulted in structures that
belonged to the two sets. The structural ambiguity therefore results
from the lack of several constraints in the N-terminus, not from the
one dihedral constraint only. Whether the N-terminus actually oc-
cupies the space on both sides of the P2-strand cannot be settled at
In a Ramachandran plot, Y5 (4 - 60", 1+4 - 20") of the wild-type
resides outside the regions typically found in proteins (Kraulis
et al., 1989). This is caused by a steric strain between the Cp atom
of Y5 and the carbonyl oxygen of H4. There must, of course, be
favorable interactions that compensate for the local unfavorable
Table 1. I : Average atomic rms deviation o f individual structures for the mean structure of backbone and all atoms.
2: Average number o f violated distance constraints for interresidue contacts. 3: Average total computed energy
Y5A" Y5Ab Y31A Y32A
I. Backbone atoms (A)
AII atoms (A)
2. Inter residue (>0.4 A)
3. Energy (kcal/mol)
2.2 k 0.81
2.8 k 0.73
1 & 0.6
60 * 8
2.0 f 0.83
2.5 f 0.76
1 f 0.6
74 f 12
0.53 k 0.15
1.0 f 0.13
1 k 0.5
23 f 13
0.81 f 0.21
1.3 f 0.26
1 k 0.5
24 f I
0.33 f 0.04
The ensemble of the the 22a and 14b final simulated annealing structures of Y5A. Correspondingly the number of structures were
18 for Y31A and 19 for Y32A. For all engineered peptides structures were refined when disulphide bridges were closed (Cys 8 and
Cys 25; Cys 19 and Cys 35). The backbone atoms (N, C", C, 0) of the residues 7-15 and 25-36 were superimposed. Values for the
wild-type from Kraulis et al., 1989.
M.-L Mattinen et al.
16 21 26 31 36 1 6
21 26 31 36 1 6 1 1 16 21 26 31
Fig. 4. All atom rms deviations of the individual structures from the average structure, and the corresponding NOEs per residue. Gray
columns represent the interresidue NOEs and black columns the intraresidue NOEs.
structures of Y31A, and (C) 19 structures of Y32A. The secondary structure elements, t y p e of +.urns (light gray sticks) and /3-strands
(dark gray arrows) are indicated at the upper part of the NOE figure of Y5A.
(A) 37 (14 + 23) structures of Y5A, (B) 18
configuration. The aromatic rings of tyrosine 5 and histidine 4 are
in parallel orientation, and it is plausible that their interaction
stabilizes the unusual +-angle
of Y5 (Loewenthal et al., 1992).
There are most likely other reasons as well, because H4 is not a
conserved amino acid (Linder et al., 1995a). It
point out that CBDEGl and CBDam have a third disulphide bridge
is of interest to
Fig. 5. Side view of the flat face. The backbones of engineered peptides (blue) are superimposed on the wild type (black) and the
spatial dispersion of the residues on the flat face are visible. Asparagine (N29) and glutamine (434) are indicated in blue. msines
(Y5, Y31, Y32) are shown in green, and the alanines in red. (A) 23 and (B) 14 structures of Y5A, (C) 18 ofY31A. and (D) 19 of Y32A.
Structure-activity relationship of engineered CBDcBHI
Fig. 6. Stereo side view of the flat face. The backbones of engineered peptides (blue) are superimposed on the wild type (black). For
clarity side chains are shown only for five structures. Asparagine (N29) and glutamine (Q34) are indicated in blue and tyrosines (Y5,
Y31, Y32) in green. The alanines are shown in r e d . (A) and (B) Y5A, (C) Y31A. and (D) Y32A.
from the tip of the N-terminus to the loop (14-19) region in the
middle of the sequence. Although this third disulphide is not con-
served, its presence in some of the CBDs may reflect the vulner-
ability of the N-terminal fold.
For Y31A and Y32A, no tp-angle restraint of YS could be de-
termined, because no NH-CaH cross-peaks were observed in the
COSY spectra. This may well be a result of the cancellation of
antiphase lines due to a small spin coupling. Hence,
the wild-type structure. In the Ramachandran plot, Y5 (4 - 60", +
- 70") of Y31A and (4 - loo", $ - 80") of Y32A are close to the
region of Y5 in the wild type even though no +-angle restraints
were used. For
YSA the corresponding cross-peak was visible
(3.Jma = 6.6 Hz), and the Corresponding dihedral angles are 4 -
4 = 60" as in
-60", $ - 120" and 4 - -loo", $ - - 170" for the best structures
in the two conformational families. These angles of AS are more
conventional and different from the backbone dihedrals for YS of
the wild type. This indicates that primarily for steric reasons, a
bulky and rigid aromatic ring at position five is required to main-
tain the unusual backbone angles and stabile fold
in the wild-type
The slight difference in pH between the binding studies (Reini-
kainen et al., 1995) and the structural studies has been shown to
of no importance for correlating structural changes with binding
affinities (Linder et al., 1995a). The removal of the aromatic ring
Mattinen et al.
at the tip of the CBD causes no major structural consequences on
the remaining residues of the flat face. We suggest that this is the
explanation for the small but detectable binding affinity of Y31A
to cellulose (Linder et al., 1995a). Because the structural changes
are small, the aromatic ring of Y31 has an important contribution
to the binding. In the case of Y32A, the local structural changes are
inevitable, because removal of the aromatic ring would without
sidechain rearrangements leave the interior of CBD exposed. The
reorganization of the side chains on the flat face thus causes a com-
plete loss of affinity towards cellulose. The aromatic ring of Y32
undoubtedly contributes to the binding affiity, but the mutation to
an alanine also affects other functional residues. In the case of Y5A,
it is difficult to partition the role of Y5 into structural or functional.
Part of the backbone folding is destroyed and also the remaining struc-
tural framework is subject to smaller changes. Consequently, it is
not surprising that there is no affiinity towards cellulose.
Based on these observations we propose that
is dictated by the precise positions of the residues on the flat face.
The role of this face in the binding of CBD to the surface of
cellulose should now be examined by specific substitutions of the
tyrosines, for example, by phenylalanines, to reveal the role of
hydroxyls, and by histidines and tryptophanes (except W
which is in the spatially restricted position)
the hydrophobicity of the aromatic rings. In previous work the
binding strength of the Y31H mutant of the CBHI enzyme has
been found to be 60% of that for the wild type (Reinikainen et al.,
1995). CBDEGI with W5 has roughly twice the affiity for cellu-
lose as CBDCsHl with Y5 (Linder et al., 1995b). Attempts to en-
gineer more strongly binding CBDs could focus on designing larger
flat faces. This approach has been exploited by combining two
CBDs by a linker peptide (our unpublished results).
In the flat face of type I CBD the periodicity of the aromatic
rings and amides and the periodicity
the polymers of the crystalline cellulose appear complementary
(Fig. 7). Four side chains of the wild-type peptide, Y5, Y31, Y32,
and 434, can be aligned well along the cello-oligosaccharide chain.
The alignment is within the precision of structure determination,
i.e., 0.52 A (Kraulis et al., 1989). A face-to-face stacking of aro-
matic rings of amino acids with sugar rings and extensive hydro-
gen bonding of polar planar groups of amino acid side chains with
sugar hydroxyls have been observed in other cases (Quiocho, 1986).
The hydrophobic effect has been proposed to account for the im-
proved affinity at increased ionic strength (Reinikainen
Structural motifs for these interactions have also been found, for
example, in the hydrolytic sites of the core domains of CBHI
(Divine et al., 1994) and CBHII (Rouvinen et al., 1990), as well as
in the type I1 CBD of Cex (Xu et al., 1995).
At the abundant 110 and 1-10 surfaces of crystalline cellulose
the sugar polymers stack stepwise on
that only parts of the rings are exposed. It is only the obtuse
comer of the cellulose crystal that
rings (Chanzy et al., 1984; Reinikainen et al., 1995), which would
be a prerequisite in the binding model. However, there is exper-
imental evidence that most of the surface of a cellulose crystal
is covered by CBD at saturation (Linder et al., 1995b). But it is
unclear if CBDs bound on 110 or
tribute to the crystal breakdown. Face-to-face stacking
matic rings to the sugar rings may be required
binding. The obtuse comer of the intact cellulose is prone to
attack. When the polymer of the obtuse comer becomes solv-
ated, more (020) surface is exposed.
the binding strength
to enlighten the role of
of the glucose rings along
et al., 1995).
the top of each other so
has fully exposed glucose
1-10 surfaces actually con-
Fig. 7. Engineered peptides (A-D) and wild type CBDCBHI placed above
the cello-oligosaccharide with aromatic rings stacked
Only the residues Y5, N29, Y31, Y32,
color coding is conventional except i n the sugar rings carbons and hydro-
gens are gray.
on top of sugar rings.
and 434 are shown for clarity. The
Based on the results presently presented,
Y31 and Y32 have a functional role for binding to the surface of
crystalline cellulose. Their structural role for backbone folding is
minor. Y5 has pronounced importance in the structural integrity of
the N-terminus of the peptide and probably also functional impor-
tance via contributions to the binding affinity. We have thus been
able to point out functionally important amino acid side chains,
knowledge that can be used in the future for modeling studies of
cellulose-cellulase interaction and in designing CBDs with altered
it is concluded that
Materials and methods
The synthesis of the three engineered peptides (Y5A, Y31A, and
Y32A) and the chemical composition of the purified peptides have
been described previously (Linder et al., 1995a). Samples for the
NMR experiments were prepared by dissolving 5.0 mg (Y5A), 5.6
Structure-activity relationship o f engineered CBDcsH,
mg (Y31A), or 6.5 mg (Y32A) of the peptide in 'H20 containing
about 10% 2H20 for the lock signal. In order to slow down the
exchange rate of the amide protons, the pH of each solution was
adjusted to 3.9. This same procedure was used in the structure
determination of the wild type (Kraulis et al., 1989). NaN3(0.02%)
was added to the samples as an anti-microbial agent. Final sample
volume was 0.7 mL.
CD measurements were carried out on a Jasco 5-710/720 spectro-
polarimeter. The samples were diluted from the NMR samples
with a 50 pM acetate buffer (pH = 5.0) to obtain peptide con-
centrations in the range 1-20 pM. Molar ellipticities 0 (mdeg)
were measured in a 1 mm path length cell as a function of wave-
length from 190-250 nm.
All experiments were performed with a Varian Unity 600 spec-
trometer. The experiments were conducted at 15 "C, and certain
spectra were also acquired at 1 "C and 25 "C. COSY (Marion &
Wiithrich, 1983), RELAY-COSY (7 = 35 ms) (Wagner, 1983),
NOESY (mixing times: 40, 80, 120, 150 or 160, and 200 ms)
(Macura & Ernst, 1980), and sensitivity enhanced TOCSY (mixing
times: 30, 50, 80, and 120 ms) (Braunschweiler & Ernst, 1983;
Bax & Davis, 1985; Cavanagh & Rance, 1990) were recorded with
standard pulse sequences and phase cycling. In addition, NOESY
spectra with 300 ms mixing time at I "C were measured for Y3 1 A
and Y32A. The FIDs were recorded with 2048 complex data points.
The number of tl increments varied, but usually it was 512 in
COSY, 400 in RELAY-COSY, 320 in NOESY, and 256 in TOCSY
with 128 or 64 transients per increment. The spectral widths were
6400 Hz in both dimensions. The solvent resonance (4.88 ppm at
15 "C) was suppressed by low-power continuous irradiation during
the relaxation delay (1.2 s). The residual signal was further reduced
by a time-domain data deconvolution method using a sinebell
filter with 20-40 points (Marion et al., 1989; Sodano & Delepi-
erre, 1993). For COSY and RELAY-COSY spectra sinebell weight
functions and for NOESY and TOCSY spectra shifted sinebell
square weight functions were used in both dimensions. The final
data matrices consisted of 2 X Ik real points. The data were
Fourier transformed, displayed, and assigned with the Felix data
processing and analysis program (Biosym Technologies, San Di-
ego, versions 2.1 and 2.3).
The assignment strategy for 'H spin systems of different amino
acids was based on the amide proton chemical shift (Chazin et al.,
1988). The assignment was partially facilitated by comparisons
with the chemical shifts of the wild type (Kraulis et al., 1989), but
the chemical shifts of the residues adjacent to the mutations dif-
fered too much from the corresponding residues of the wild type to
be deduced by comparison. This was also the case for some other
residues, especially in the N-termini of the engineered peptides.
For these amino acids the assignments were inferred by applying
the strategy described by Billeter et al. (1982) and Chazin and
Wright (1987). In the case of Y31A, the assignment of the shifts
for most protons was obtained from a comparison with the wild
type, because residue 3 1, located at the tip of the molecule, in-
fluenced the chemical shifts of the neighboring residues only
moderately. For Y5A and Y32A, assignment was somewhat less
straightforward, as the ring current effect of the tyrosines affected
resonance frequencies of many adjacent protons. The complete
sequence and stereo-specific assignments were obtained for all
mutants. The wild-type-like pairing of the two disulphide bridges
was also confirmed for the engineered peptides from the NOES of
CPH of the cysteines.
For every unambiguously assigned and well-resolved cross-peak
the initial slope of the NOE build-up curve was calculated from a
fitted second-order polynomial and converted into distances by
calibrating the initial slope to distance. Covalent distances between
certain geminal protons were used for the calibration (1.78 A). The
determined distances were given with *20% uncertainty ranges.
Intensities of protons with degenerate shifts were divided by the
number of equivalent protons (e.g., by three in the case of a methyl
group) before the distances were extracted and an extra 0.5 8, was
added to obtain the upper bounds (Clore et al., 1987; Wagner et al.,
1987). Distances corresponding to weak or poorly resolved cross-
peaks were given lower bounds of at 1.8 A, and upper bounds
6.0 A. For Y31A and Y32A, more cross-peaks were identified
from the NOESY spectra recorded at 1 "C and with a 300 ms
mixing time. The restraint sets were supplemented with these qual-
itative restraints with lower and upper bounds of I .8 A and 5.0 A,
respectively, and extended by 1 A for pseudo-atoms.
3JHN,and 'JeP coupling constants were measured from COSY
spectra in 'H20 by the J-doubling method (Le Parco et al., 1992;
McIntyre & Freeman, 1992). For Y32A, 'Jmp coupling constants
had to be determined from the spectrum recorded in 2H20, as the
C"H-CPH region was disturbed by the residual water signal. Ste-
reospecific assignments of the 0-methylene protons were obtained
based on NOES if one of the 'JaP was greater than 10 Hz or both
were smaller than 5 Hz (Wagner et al., 1987). The 3JHNa and 3Jap
coupling constants were converted to 4- and X-angles by means of
Karplus relations (deMarco et al., 1978; Smith et al., 1996). The
final dihedral angle restraints were obtained by adding a +30°
margin to the determined dihedral angle.
The number of distance restraints imposed during calculations and
the corresponding values for the wild type (Kraulis et al., 1989) are
summarized in Table 2. The total number of distance restraints was
524 for Y31A and 519 for Y32A. For Y5A only 295 distance
restraints could be identified. This mutant also had significantly
fewer dihedral angle restraints (32) than Y31A (42), Y32A (42),
and the wild type (82). The final set of structures was computed
with a total of 327 experimental restraints for Y5A, 566 for Y3 IA,
and 561 for Y32A. The corresponding value for the wild type was
660, which included 24 hydrogen bond restraints imposed after
structure generation assuming certain acceptor oxygens, and 24
*-angle restraints obtained by modeling. These were not used in
our structure determinations. Therefore, in our calculations the
number of experimental restraints for Y31A and Y32A were com-
parable to wild type. For Y5A, the number of experimental re-
straints was only about 60% of the restraints for the other peptides.
Initially ensembles of 60 (Y5A), 40 (Y31A), and 40 (Y32A)
structures were generated by the standard protocols of the distance
M.-L. Muttinen et al.
Table 2. Number of experimental restraints used for structure culculutions
Peptide ISPA Intensity Qualitative tal
Interresidue, short range ( Ii -
Interresidue, long range ( / i -
Interresidue, short range (li - j l ) 5 5
Interresidue, long range ( 1 i - jl) > 5
Interresidue, short range (]i - j l ) 5 5
Interresidue, long range ( ( i - j ( ) > 5
Interresidue, short range (li -
Interresidue, long range (li -
0 295 22 10 0 0 327
I27 524 24
0 0 566
25 519 24
25 24 24 660
aResidues i and j belong to different spin systems.
'Values obtained from Kraulis et al., 1989.
geometry (DGII, Discover 3. I, Biosym Technologies, San Diego,
CA) (Crippen, 1977; Havel et al., 1983, 1990, 1991; Sheeck et al.,
1991). Because Y5A had fewer restraints than the other engineered
peptides, it was necessary to map the conformation space with a
larger number of structures. The stereospecific assignments of gly-
cine a protons were deduced after the initial structure generation
and later imposed. Upon inspection of the initial structures new
cross-peaks were identified. The DGII calculations were then re-
peated and the structures refined by simulated annealing (SA, Dis-
cover 3.1, Biosym Technologies, San Diego, CA) (Clore et al.,
1986; Nilges et al., 1988a, 1988b, 1988~). In the molecular dy-
namics (MD) computation, the AMBER force field was used.
SAlMD structures were finally minimized by a conjugate gradi-
ents algorithm. The structures were inspected with the Insight11
(Biosym Technologies, San Diego, CA) molecular modeling soft-
ware to detect inconsistencies and restraint violations. All calcu-
lations and graphical analyses were performed on Silicon Graphics
IRIS 4D30 TG and Indigo 2 workstations.
The co-ordinates for
Brookhaven Protein Data Bank and the chemical shifts in the BioMagRes-
Bank at Madison. This work has been supported by the Academy of Fin-
land, the Emil Aaltonen Foundation, the foundation of
Ehrnrooth, the Swedish National Board for Industrial and Technical De-
velopment, and the Neste Research Foundation.
YSA, Y31A, and Y32A will be deposited
Ella and Georg
Bax A, Davis DG. 1985. MLEV-17 based two-dimensional homonuclear mag-
netisation transfer spectroscopy. J Magn Reson 65:355-360.
Btguin P, Aubert PJ. 1994. The biological degradation of cellulose. FEMS
Microbiol Rev 13:25-58.
Billeter M, Braun W, Wiithrich K. 1982. Sequential resonance assignments in
protein proton nuclear magnetic resonance spectra. Computation of steri-
cally allowed proton-proton distances and statistical analysis of proton-
proton distances in single crystal protein conformations. J Mol Biol 155:321-
Braunschweiler L, Ernst RR. 1983. Coherence transfer by isotropic mixing:
Application to proton correlation spectroscopy. J Magn Reson 53:521-528.
Cambillau C, van Tilbeurgh H. 1993. Structure of hydrolases: Lipases and
cellulases. Curr Opin Struct Biol 3%-895.
Cavanagh J, Rance M. 1990. Sensitivity improvement in isotropic mixing
(TOCSY) experiments. J Magn Reson 8872-85.
Chanzy H. Henrissat B, Vuong R, Schiilein M. 1984. The action of 1.4-p.~-
glucan cellobiohydrolase on Valonia cellulose microcrystals. An electron
microscopic study. FEES Left 153:113-118.
Chazin WJ, Wright PE. 1987. A modified strategy for identification of proton
spin systems in proteins. Biopolymers 26:973-977.
Chazin WJ, Rance M, Wright PE. 1988. Complete assignment of the 'H nuclear
magnetic resonance spectrum of french bean plastocyanin. Application of an
integrated approach to spin system identification in proteins. J Mol Biol
Claeyssens M, Tomme P. 1989. Structure-function relationships of cellulolytic
proteins from Trichoderma reesei. In: Kubicek CP, ed. Trichoderma reesei
cellulases: Biochemistry, genetics, physiology and application. Cambridge:
Royal Society of Chemistry. pp 1-1 I.
Clore GM, Briinger AT, Karplus M, Gronenborn AM.
molecular dynamics with interproton distance restraints to three-dimensional
protein structure determination: A model study of crambin. J Mol Bid
Clore GM, Gronenborn AM, Nilges M. Ryan CA. 1987. The three-dimensional
structure of potato carboxypeptidase inhibitor in solution: A study using
nuclear magnetic resonance, distance geometry and restrained molecular
1986. Application of
dynamics. Biochemisfry 2623012-8023.
Crippen GM. 1977. A novel approach to the calculation of conformation: Dis-
tance geometry. J Comp Phys 2496-107.
deMarco A, Linas M, Wiithrich K. 1978. Analysis of the 'H-NMR spectra of
femchrome peptides. 1. The non-amide protons. Biopolymers 17617-636.
Din N. Forsythe IJ, Burtnick LD, Gilkes NR, Miller RC, Warren RAJ. Kilburn
DG. 1994. The cellulose-binding domain of endoglucanase A (CenA) from
Cellumonasjimi: Evidence for involvement of tryptophan residues in bind-
ing. Mol Microbiol 11:747-755.
Divine C, Stihlberg J, Reinikainen T, Ruohonen L, Pettersson G. Knowles JKC.
Teeri TT, Jones TA. 1994. The three-dimensional structure of cellobiobydro-
lase I from Trichoderma reesei reveals a lysozyme-like active site on a
lecitin-like framework. Science 265524-528.
Gardner KH, Blackwell I. 1974. The structure of native cellulose. Biopolymers
Havel TF. 1990. The sampling properties of some distance geometry algorithms
Structure-activity relationship of engineered CBDcsHl
applied to unconstrained polypeptide chains: A study of 1830 independently
computed conformations. Biopolymers 2 9 1565-1585.
Havel TF. 1991. An evaluation of computational strategies for use in the deter-
mination of protein structure from distance constraints obtained by nuclear
magnetic resonance. Prog Biophys Mol Biol 5643-78.
Havel TF, Kuntz ID, Crippen GM. 1983. The theory and practice of distance
geometry. Bull Math Biol 45:665-720.
Henrissat B, Driguez H, Viet C, Schiilein M. 1985. Synergism of cellulases from
Trichoderma reesei in the degradation of cellulose. Bio/Technology 3:722-
Irwin DC, Spezio M, Walker LP, Wilson DB. 1993. Activity studies of eight
purified cellulases: Specificity, synergism and binding domain efforts. Bio-
techno/ Bioeng 42:1002-1013.
Kraulis PJ, Clore MG, Nilges M, Jones AT, Pettersson G, Knowles J, Gronen-
born AM. 1989. Determination of the three-dimensional structure of the
C-terminal domain of cellobiohydrolase I from Trichoderma reesei. Bio-
Le Parco JM, McIntyre L, Freeman R. 1992. Accurate coupling constants from
two-dimensional correlation spectra by “I deconvulsion.” J Magn Reson
Linder M, Mattinen ML, Kontteli M, Lindeberg G, Stihlberg J, Drakenberg T,
Reinikainen T, Pettersson G, Annila A. 1995a. Identification of functionally
important amino acids in the cellulose-binding domain of Trichoderma re-
esei cellobiohydrolase I. Protein Sei 4: 1056-1064.
Linder M, Lindeberg
G, Reinikainen T, Teen T, Pettersson G. 1995b. The
difference in affinity between two fungal cellulose-binding domains is dom-
inated by a single amino acid substitution. FEBS Lett 372:96-98.
Loewenthal R, Sancho J, Fersht AR. 1992. Histidine-aromatic interactions in
bamase. J Mol Biol 224:759-770.
Macura A, Emst RR. 1980. Elucidation of cross relaxation in liquids by two-
dimensional NMR spectroscopy. Mol Phys 419-1 17.
Marion D, Wiithrich K. 1983. Application of phase sensitive two-dimensional
correlated spectroscopy (COSY) for measurements of ‘H-’H spin-spin cou-
pling constants. Biochem Biophys Res Commun 113:967-974.
Marion D, Mitsuhiko I, Bax A. 1989. Improved solvent suppression in one- and
two-dimensional NMR spectra by convolution of time-domain data. J Magn
Mclntyre L, Freeman R. 1992. Accurate measurement of coupling constants by
J doubling. J Magn Reson 96:425-431.
Nidetzky B, Hayn M, Macarron R, Steiner W. 1993. Synergism of Trichoderma
reesei cellulases while degrading different celluloses. Biotechnol Lett 15:71-
Nilges M, Clore GM, Gronenbom AM. 1988a. Determination of three-dimensional
structures of proteins from interproton distance data by hybrid distance
geometry-dynamical simulated annealing calculations. FEBS Lett 229317-
Nilges M, Gronenbom AM. Briinger AT, Clore GM. 1988b. Determination of
three-dimensional structures of proteins by simulated annealing with inter-
proton distance restraints: Application to crambin, potato carboxypeptidase
inhibitor and barley serine proteinase inhibitor 2. Protein Eng 2:27-38.
Nilges M, Clore GM, Gronenbom AM. 1988~. Determination of three-dimensional
structures of proteins from interproton distance data by dynamical simulated
annealing from a random array of atoms. FEBS Lett 239: 129-136.
Poole DM, Hazlewood GP, Huskisson NS, Virden R, Gilbert HJ. 1993. The role
of conserved tryptophan residue in the interaction of a bacterial cellulose
binding domain with its ligand. FEMS Microbiol Lett 106:77-84.
Quiocho FA. 1986. Carbohydrate-binding proteins: Tertiary structures and protein-
sugar interactions. Annu Rev Biochem 55:287-315.
Quiocho FA. 1993. Probing the atomic interactions between proteins and car-
bohydrates. Biochem SOC Transact 21:442-448.
Reinikainen T, Ruohonen L, Nevanen T, Laaksonen L, Kraulis P, Jones TA,
Knowles JKC, Teen T. 1992. Investigation of the function of mutated
cellulose-binding domains of Trichoderma reesei cellobiohydrolase 1. Pro-
teins Struct Funct Genet 14:475-482.
Reinikainen T, Teleman 0, Teen ’IT. 1995. Effects of pH and high ionic strength
on the adsorption and activity of native and mutated cellobiohydrolase I
from Trichoderma reesei. Proteins 22:392-403.
Rouvinen J, Bergfors T, Teeri T, Knowles JKC, Jones
dimensional structure of cellobiohydrolase I1 from Trichoderma reesei. Sci-
Sheeck RM, Torda AE, Kemmink J, van Gunsteren WE 1991. Structure deter-
mination by NMR: The modelling of NMR parameters as ensemble aver-
ages. In: Hoch JC, Poulsen FM, Redfield C, eds. Computational aspects of
the study o f biological macromolecules by nuclear magnetic
spectroscopy. New York: Plenum Press. pp 209-217.
TA. 1990. Three-
Smith LJ, Bolin KA, Schwalbe H, MacArthur MW, Thomton JM, Dobson CM.
1996. Analysis of main chain torsion angles in proteins: Prediction of NMR
coupling constants for native and random coil conformations. J Mol Biol
Sodano P, Delepierre M. 1993. Clean and efficient suppression of the water
signal in multidimensional NMR spectra. J Magn Reson 104:88-92.
Stihlberg J, Johansson G, Pettersson G. 1988. A binding-site-deficient, catalyt-
ically active, core protein of endoglucanase 111 from the culture filtrate of
Trichoderma reesei. Eur J Biochem 173:119-183.
Stihlberg J, Johansson G, Pettersson G. 1991. A new model for enzymatic
hydrolysis of cellulose based on the two-domain structure of cellobiohydro-
lase 1. Bio/Technology 9:286-290.
Tomme P, Tilbeurgh H van, Pettersson G, Damme J van, Vandekerckhove J,
Knowles J, Teen T, Claeyssens M. 1988. Studies of the cellulolytic system
of Trichoderma reesei QM 9414. Analysis of domain function in two cello-
biohydrolases by limited proteolysis. Eur J Biochem 170575-581.
van Tilbeurgh H, Tornme P, Claeyssens M, Bhikhabhai R, Pettersson G. 1986.
Limited proteolysis of the cellobiohydrolase I from Trichoderma reesei.
FEBS Lett 204:223-227.
Vyas NK. 1991. Atomic features of protein-carbohydrate interactions. Curr
Opin Struct Biol 1:723-740.
Wagner G. 1983. Two-dimensional relayed coherence transfer spectroscopy of
a protein. J Magn Reson 55:151-156.
Wagner G, Braun W, Havel TF, Schaumann T, Go N, Wiithrich K. 1987. Protein
structures in solution by nuclear magnetic resonance and distance geometry.
The polypeptide fold of the basic pancreatic trypsin inhibitor determined using
two different algorithms, DISGEO and DISMAN. J Mol Biol 196:611-639.
Wood TM, Garcia-Campayo V. 1990. Enzymology of cellulose degradation.
Biodegradation I : 147- I6 1,
Xu CY, Ong E, Gilkes NR, Kilbum DG, Muhandiram DR, Hams-Brandts M,
Carver JP, Kay LE, Havery TS. 1995. Solution structure of a cellulose-
binding domain from Cellumonasfimi by nuclear magnetic resonance spec-
troscopy. Biochemist. 34:6993-7009.