Journal of Molecular Catalysis B: Enzymatic 48 (2007) 84–89
Influence of self-assembled monolayer surface chemistry on
Candida antarctica lipase B adsorption and specific activity夽
Joseph A. Laszlo ∗ , Kervin O. Evans
New Crops and Processing Technology Research, USDA-Agricultural Research Service, National Center for Agricultural Utilization Research,
1815 N. University Street, Peoria, IL 61604, USA
Received 24 May 2007; received in revised form 10 June 2007; accepted 22 June 2007
Available online 1 July 2007
Abstract
Immobilization of Candida antarctica B lipase was examined on gold surfaces modified with either methyl- or hydroxyl-terminated selfassembled alkylthiol monolayers (SAMs), representing hydrophobic and hydrophilic surfaces, respectively. Lipase adsorption was monitored
gravimetrically using a quartz crystal microbalance. Lipase activity was determined colorimetrically by following p-nitrophenol propionate hydrolysis. Adsorbed lipase topography was examined by atomic force microscopy. The extent of lipase adsorption was nearly identical on either surface
(approximately 240 ng cm−2 ), but its specific activity was sixfold higher on the methyl-terminated SAM, showing no activity loss upon immobilization. A uniform, 5.5 nm high, highly packed monolayer of CALB formed on the methyl-terminated SAM, while the adsorbed protein was
disordered on the hydroxyl-terminated SAM. Hydrophobic surfaces thus may specifically orient the lipase in a highly active state.
Published by Elsevier B.V.
Keywords: Protein adsorption; Lipase; Immobilization
1. Introduction
Nonaqueous biocatalysis is finding commercial utility in
the production of fine and specialty chemicals [1–3]. Lipases
are among the most broadly deployed biocatalysts because of
their ability to produce chiral chemicals with high enantiomeric
purity [4]. Lipases catalyze hydrolysis, alcoholysis, esterification and transesterification of carboxylic acids or esters. The
B lipase from Candida antarctica (CALB) has been a synthesis
workhorse, as well as the subject of numerous fundamental studies regarding nonaqueous enzymology [5–7]. CALB typically
is used in an immobilized form, such as the commercial product
Novozym 435.
Enzyme immobilization offers many potential benefits.
Immobilization can improve enzyme operational performance
and stability, as well as provide for ready separation of bio-
夽 Names are necessary to report factually on available data; however, the
USDA neither guarantees nor warrants the standard of the product, and the use
of the name by the USDA implies no approval of the product to the exclusion
of others that may also be suitable.
∗ Corresponding author. Tel.: +1 309 681 6322; fax: +1 309 681 6686.
E-mail address: Joe.Laszlo@ars.usda.gov (J.A. Laszlo).
1381-1177/$ – see front matter. Published by Elsevier B.V.
doi:10.1016/j.molcatb.2007.06.010
catalyst from the reaction medium [8,9]. CALB has been
immobilized on various support materials such as porous resins
and silicas [5,10–15]. Polypropylene and acrylic resins, regarded
as hydrophobic supports, are particularly efficacious. These
meso- and macroporous polymeric materials accommodate high
CALB loadings (up to 20% w/w protein) and good retention of
enzyme activity. It is not understood how the support matrix
influences CALB activity. The support matrix can impose activity limitations by altering the lipase’s native conformation or by
diminishing substrate diffusivity. Distinguishing between these
influences can be difficult. Infrared detection of CALB on various supports indicates a very heterogenous distribution within
the polymer matrix [12,16]. Potentially, there are substantial
amounts of lipase that do not interact directly with the polymer surface. Thus, conventional immobilization materials do
not provide a sufficiently uniform enzyme–support interface
for study of support surface influences on CALB activity and
topography.
The current study examines the impact of support surface properties (hydrophobicity) on CALB activity, in an
arrangement that does not impose substrate internal transport
limitations, through the use of flat, self-assembled alkylthiolmodified gold surfaces [17]. A correlation between surface
J.A. Laszlo, K.O. Evans / Journal of Molecular Catalysis B: Enzymatic 48 (2007) 84–89
hydrophobicity and enzyme topographical arrangement is evaluated.
2. Experimental
2.1. Reagents and materials
CALB in solution (commercial name Lipozyme CALB-L)
was obtained from Novozymes North America. CALB in the
form of a highly purified enzyme powder (stated purity 95%) was
purchased from Polium Technologies (Hoffman Estates, Illinois,
USA), as was p-nitrophenylpropionate (pNPP). The powder
form of CALB was used solely in the preparation of protein
calibration standards for determining the CALB concentration
in the liquid sample. 1-Undecanethiol (MU), 11-mercapto-1undecanol (MUOH), and ethanol were from Sigma–Aldrich.
Water was obtained from a Barnstead NANOpure Diamond UV
ultrapure water purification system (resistivity 18.2 M cm).
2.2. Self-assembled monolayer (SAM) preparation
The Au surfaces of QCM sensor crystals (Q-Sense) and
Au-coated glass slides (Platypus Technologies) were cleaned
sequentially by UV/O3 treatment (15 min), H2 O/NH4 OH/H2 O2
(5:1:1 v/v/v) at 70 ◦ C (15 min), another UV/O3 treatment
(15 min), and finished with rinses of water and then ethanol.
Au-covered mica (SPI Supplies) was not subjected to the
H2 O/NH4 OH/H2 O2 cleaning step but instead was given water
and methanol rinses in between UV/O3 treatments. Cleaned
substrates were immersed for at least 24 h in ethanolic solutions of 10 mM MU or MUOH, respectively forming methylor hydroxyl-terminated SAMs. SAM surfaces were rinsed with
ethanol and dried under a stream of N2 . Static water contact
angles on SAM-modified Au-coated glass slides were measured immediately using a FTA200 optical system (First Ten
Angstroms Inc., Portsmouth, VA).
2.3. Protein and hydrolytic activity assays
CALB concentration in the solution received from
Novozymes was determined by the bicinchoninic acid method
[18,19], using purified CALB powder to prepare calibration
standards. Samples and standards were incubated at 37 ◦ C for
30 min, allowed to cool to room temperature, and then their
absorbances were measured at 562 nm.
CALB specific activity was assessed by following
the catalytic generation of the p-nitrophenolate anion
(15,000 M−1 cm−1 at 410 nm) from pNPP hydrolysis [20]. One
unit (U) of lipase activity produces 1 mol of product per min.
Reactions were conducted in 10 mM KHPO4 , pH 7.0, buffer
containing 1.0 mM pNPP at 25 ◦ C. Buffer-solubilized CALB
activity was determined in 1-cm-pathlength cuvettes using a
Shimadzu 1240 UV–vis spectrophotometer with the enzyme
at a concentration of 4.5 g/mL. Color development was linear with time for several minutes. pNPP hydrolysis in the
absence of CALB was negligible. To determine the activity
of surface-immobilized CALB, SAM-modified Au-coated glass
85
slides (25 mm × 75 mm) were incubated for 3 h at room temperature in buffer containing 450 g/mL (13.5 M) CALB. Slides
were transferred to buffer for 2 min to remove loosely adhered
protein, then briefly rinsed with a stream of water. Excess water
was wicked from the surface. To create reaction wells on the
slides, flat-sided glass O-rings (2.25 cm i.d. and 0.5 cm high)
were attached to the slides (two O-rings per slide) with vacuum grease. Each O-ring enclosed 4.0 cm2 of slide surface. The
constructed wells were filled with 1.0 mL of buffer containing 1.0 mM pNPP and the slides were placed in a forced-air
orbital shaker operating at 90 rpm and 25 ◦ C. Slides were covered with a Petri dish to minimize fluid evaporation from the
wells. At timed intervals, well reaction medium (1.0 mL) was
transferred to cuvettes for photometric analysis of the reaction
product concentration. Because the reaction with immobilized
CALB evolved more slowly than with CALB in solution, a slight
amount of uncatalyzed product formation was observed to occur.
Therefore, a control solution of pNPP was used to subtract background, uncatalyzed nitrophenolate anion generation from the
CALB-catalyzed reaction. The uncatalyzed rate was approximately 7% of that of the slowest catalyzed reaction, i.e., with
CALB immobilized on a hydroxyl-terminated SAM (see Section 3.3). Four time points per reaction were taken (i.e., using
two slides), typically ranging from 2 to 20 min contact time of
the reaction buffer with the slide surface. No pNPP hydrolysis (other than at the uncatalyzed hydrolysis rate) was observed
with SAM-modified surfaces lacking adsorbed CALB. Analysis
of CALB activity on each SAM surface (methyl- and hydroxylterminated) was performed four to six times.
2.4. Quartz crystal microbalance (QCM) measurements
QCM measurements of CALB adsorption to SAM-modified
Au surfaces on AT-cut quartz crystals were performed with a
Q-Sense D300 system (Q-Sense Inc., Glen Burnie, MD). The
crystal and solution chamber temperature was maintained at
25.0 ◦ C. The QCM technique, described in detail elsewhere
[21,22], provides information about the amount of adsorbed
mass through changes in vibrational frequency (f). For rigid
films, such as that obtained with the adsorbate dried onto the
crystal surface (see below), the Sauerbrey equation [21] can be
employed to determine adsorbed mass (m):
C
m = −
f
(1)
N
where C is the mass sensitivity constant, 17.7 ng cm−2 Hz−1 the
primary harmonic (5 MHz) and N is the overtone number. The
15 MHz (N = 3) overtone was used for quantifying protein dry
mass (ng cm−2 ).
SAM-modified crystals were assembled into the QCM unit,
and then the cell was flushed 20 min with N2 . The system was
allowed to equilibrate overnight at temperature, which was necessary to eliminate drift in f. The cell was flushed again with N2
(20 min), and the absolute f of the third overtone was recorded.
Buffer (10 mM KHPO4 , pH 7.0, 0.2 m filtered and degassed)
was passed into the cell to establish baseline values of f in liquid.
86
J.A. Laszlo, K.O. Evans / Journal of Molecular Catalysis B: Enzymatic 48 (2007) 84–89
This was followed by a solution of CALB (450 g/mL) in buffer,
which allowed CALB to adsorb onto the crystal surface under
non-flow conditions, and f was continuously monitored in the
conventional manner. The cell was then successively flushed
with water, to remove buffer salts (no changes in f and D apparent), N2 (20 min), ethanol (5 mL) and again with N2 (20 min) to
provide a dry protein film on the crystal surface and the absolute f
(under N2 ) was recorded. The decrease in f (f) was translated to
m using Eq. (1). Preliminary experiments showed that, absent
CALB adsorption, this treatment regimen had minimal impact
on measurements in air (f < 3 Hz).
2.5. Atomic force microscopy (AFM)
CALB adsorbed to SAM-modified Au on mica was imaged
in air using tapping mode AFM with a Nanoscope IV (Veeco
Metrology) instrument equipped with a diamond-like carbon
spike (DLCS) probe (1–3 nm tip radius).
Fig. 1. Representative QCM CALB adsorption traces with methyl- and
hydroxyl-terminated SAM surfaces (solid black and dotted red lines, respectively). Frequency change (f) was monitored at the crystal’s third overtone
(approximately 15 MHz).
2.6. Molecular modeling
The CALB crystal structure [23] 1TCA was downloaded
from the Protein Data Bank. RasTop version 2.0.3 was used
for rendering and intramolecular distance measurement.
3. Results and discussion
3.1. Surface characterization
The influence of surface hydrophobicity on CALB adsorption and activity was examined. Self-assembly of alkane thiols
on Au surface provides a method of forming surfaces with
uniform properties. Water contact angle measurements indicated that hydroxyl- and methyl-terminated SAM surfaces were
hydrophilic (20◦ ) and hydrophobic (75◦ ), respectively. This
large difference in surface hydrophobicity was considered adequate to compare its effect on CALB adsorption and activity.
3.2. QCM analysis of CALB adsorption
CALB adsorption to a methyl-terminated SAM as observed
by the QCM technique occurred in two kinetic regimes (Fig. 1).
Three-fourths of the final frequency shift occurred within the
mixing time of the cell (<2 min), which was followed by a very
slow adsorption phase extending almost to 180 min, at which
point no further adsorbed mass increase was observed. Rinsing the cell with buffer removed loosely adhered protein (small
increases in f). CALB adsorption to a hydroxyl-terminated SAM
displayed a similar QCM response (Fig. 1). With either surface, CALB showed no indication of desorbing in buffer after
the cell rinse, indicating that stable protein films were obtainable with either hydrophobic or hydrophilic surfaces. The f
values after the buffer rinse were not significantly different for
the two surfaces (110 ± 10 and 137 ± 25 Hz, respectively, for
the methyl- and hydroxyl-terminated SAM surfaces), indicating
that approximately similar amounts of mass adsorbed in each
case.
For proteins adsorbed from buffer onto a surface, the adsorbed
mass observed reflects contributions from protein and acoustically coupled water when measured by the QCM technique.
To provide a precise determination of adsorbed CALB mass
alone, the cell and crystal surface were dried and the resulting f value was determined in atmosphere. Applying the
Sauerbrey Eq. (1) to CALB adsorbed to a methyl-terminated
SAM surface following a 180 min equilibrium period indicated
an adsorbed CALB mass of 242 ± 38 ng cm−2 . The hydroxylterminated SAM had a similar amount of CALB dry mass
adsorbed to it (244 ± 115 ng cm−2 ). Therefore, the CALB dry
mass values correlated with the observed hydrated mass f values (Fig. 1). Wannerberger and Arnebrant [24] estimated closed
packed monolayer coverage would be 274 ng cm−2 for the lipase
with a 4.0 nm × 5.0 nm surface orientation. The observed extent
of CALB binding on either surface thus was consistent with near
monolayer coverage.
Commercially sourced CALB shows a single band by SDSPAGE [16,25,26], so the observed QCM response cannot be
attributed to other proteins. However, unidentified excipients in
the commercial preparation, although 100-fold diluted for these
experiments, could not be precluded as a contributing factor in
the protein’s adsorption behavior.
3.3. Immobilized CALB activity
CALB displayed dramatically different hydrolytic activities when immobilized on methyl- and hydroxyl-terminated
SAM surfaces (Fig. 2). The QCM results demonstrated that
once adsorbed to a SAM surface and buffer rinsed, no further
CALB desorption in buffer occurs. Thus, hydrolytic activity
associated with CALB detaching from the surface was not
a concern. The hydrolysis rate with CALB on the methylterminated SAM surface was sixfold higher than with it on the
hydroxyl-terminated SAM surface. Assuming 240 ng cm−2 of
CALB on either surface, the specific activities were 5.4 and
J.A. Laszlo, K.O. Evans / Journal of Molecular Catalysis B: Enzymatic 48 (2007) 84–89
Fig. 2. pNPP hydrolysis by CALB immobilized on methyl- and hydroxylterminated SAM surfaces. Methyl SAM: filled squares, solid regression line
(r2 = 0.823). Hydroxyl SAM: filled circles, broken regression line (r2 = 0.762).
0.9 U/mg for the methyl- and hydroxyl-terminated SAM surfaces, respectively. The specific activity of CALB in solution
was 4.2 U/mg, so adsorbing CALB to the methyl-terminated
SAM surface appeared to improve its activity. However, given
87
the level of uncertainty in measuring the reaction rate with
CALB immobilized, a more conservative interpretation is that
CALB did not suffer activity loss with methyl-terminated SAM
immobilization.
CALB is not known to exhibit interfacial activation, not
having a lid-like polypeptide loop in proximity to the active
site that moves in response to hydrophobic surface abutment
[23]. Blanco and coworkers [25] found that CALB adsorption
to hydrophobic silica results in substantial pNPP hydrolytic
activity loss. Reetz and coworkers [27] demonstrated only 52%
specific activity retention with CALB immobilized in hydrophobic SiO2 sol–gels. Based on a calculation of catalytically active
CALB immobilized on polypropylene, determined by active
site titration, only about 26% of the adsorbed lipase retains
activity (7.8 mg of active CALB out of the 30 mg of CALB
immobilized per g of support) [28]. Less than 50% of CALB
adsorbed to methyl methacrylate resin retains activity [16]. Thus
the observation that CALB immobilized on a methyl-terminated
SAM surface had as great or greater activity than the enzyme
in solution was somewhat unexpected. Activity retention differences between CALB on a hydrophobic resin or porous
silica and that observed here with CALB on a flat surface may
in part be ascribed to the lack of substrate internal diffusion
limitations in the latter situation, or they may reflect an archi-
Fig. 3. AFM images of CALB on (A) methyl- and (B) hydroxyl-terminated SAM surfaces. The scan size was 200 nm × 200 nm. The accompanying section image
analyses were mid-y-axis line scans. Each section analysis z-axis minimum was assigned a value of zero and all other height values are relative to this minimum.
Images were not subjected to global flattening or filtering.
88
J.A. Laszlo, K.O. Evans / Journal of Molecular Catalysis B: Enzymatic 48 (2007) 84–89
tectural arrangement of CALB on a methyl-terminated SAM
surface that does not result in adverse structural changes to the
enzyme.
Disordered CALB aggregation alone does not result in
hydrolytic activity loss. Cross-linked enzyme aggregates of
CALB retained high pNPP hydrolysis rates, some even showing
enhanced activity compared to native enzyme, depending on
the glutaraldehyde treatment conditions [29]. The substantial
loss of CALB hydrolytic activity observed upon adsorption to a
hydroxyl-terminated SAM surface therefore must be attributed
to an interfacial influence.
retention of specific activity. On this surface the protein assumed
a highly packed monolayer of uniform molecular arrangement.
CALB adsorption onto a hydrophilic, hydroxyl-terminated SAM
led to a substantially lower specific activity and disordered
protein arrangement. Interfacial interactions between the support and enzyme were responsible for the different outcomes.
Although CALB is not subject to interfacial activation, it does
have an immobilization surface preference.
Acknowledgement
We are indebted to Leslie Smith for her technical assistance.
3.4. AFM imaging of adsorbed CALB
Although CALB adsorbs to the same extent on hydroxyl- and
methyl-terminated SAM surfaces, apparently with near monolayer coverage (based on the amount of protein absorbed and
calculated surface area), the actual arrangement of the enzyme
on the two surfaces may be very different. This could account
for the large lipase activity differences between the two surfaces. AFM microscopy was employed to visualize CALB
on hydroxyl- and methyl-terminated SAM surfaces. CALB
(450 g/mL) was adsorbed from buffer solution onto atomically
flat SAM surfaces and then briefly dried under a stream of N2 .
A 10 min incubation period was selected based on the QCM-D
observation (Fig. 1) that coverage was not complete at this point
and, therefore, portions of the underlying SAM surface could be
reached with the AFM probe.
Fig. 3 shows that on either surface monolayers of CALB
were formed, although their appearance was strikingly different. On the methyl-terminated SAM, CALB molecules were
uniformly oriented with 5.5 nm peak heights. Even at very low
concentrations (4.5 g/mL), CALB formed clusters of uniform
height (not shown) although native CALB does not form aggregates in solution [30]. The 5.5 nm height found for CALB on a
methyl-terminated SAM (Fig. 3) was consistent with the CALB
crystallographic structure. Its longest intramolecular distance is
5.74 nm (Thr316 to Pro268). The observation of 5.5 nm high
CALB molecules suggests its longest axis is oriented perpendicular to the surface. CALB on a hydroxyl-terminated SAM
displayed slightly lower peak heights, 2–4 nm high, which may
reflect a flattening or distortion of the molecules on the surface
(Fig. 3B). Such surface-induced structural changes in proteins
are common [31], but there has been no such finding reported in
the literature regarding CALB. The aggregation state of CALB
on methyl methacrylate resin has been suggested recently to play
a role in the activity of CALB [16]. AFM imaging indicated
that adsorption geometry of CALB, as influenced by surface
hydrophobicity, may have an important impact on catalytic efficacy.
4. Conclusions
Consistent with the general finding that CALB immobilization on hydrophobic supports yields the best enzyme activity,
the present work demonstrated that the hydrophobic surface of
a methyl-terminated SAM allowed the lipase to adsorb with full
References
[1] A. Ballesteros, U. Bornscheuer, A. Capewell, D. Combes, J.-S. Condoret,
K. Koenig, F.N. Kolisis, A. Marty, U. Menge, T. Scheper, H. Stamatis, A.
Xenakis, Biocatal. Biotransform. 13 (1995) 1–42.
[2] Y.L. Khmelnitsky, J.O. Rich, Curr. Opin. Chem. Biol. 3 (1999) 47–
53.
[3] A. Straathof, S. Panke, A. Schmid, Curr. Opin. Biotechnol. 13 (2002)
548–556.
[4] R.D. Schmid, R. Verger, Angew. Chem. Int. Ed. 37 (1998) 1608–1633.
[5] F.J. Plou, M. Angeles Cruces, M. Ferrer, G. Fuentes, E. Pastor, M. Bernabé,
M. Christensen, F. Comelles, J.L. Parra, A. Ballesteros, J. Biotechnol. 96
(2002) 55–66.
[6] T. Kobayashi, S. Adachi, Biotechnol. Lett. 26 (2004) 1461–1468.
[7] S. Lutz, Tetrahedron: Asymmetry 15 (2004) 2743–2748.
[8] A. Bastida, P. Sabuquillo, P. Armisén, R. Fernández-Lafuente, G.
Fernández-Lorente, J.M. Guisán, Biotechnol. Bioeng. 58 (1998) 486–
493.
[9] R. Fernández-Lafuente, P. Armisén, P. Sabuquillo, G. Fernández-Lorente,
J.M. Guisán, Chem. Phys. Lipids 93 (1998) 185–197.
[10] M. Arroyo, J.M. Sánchez-Montero, J.V. Sinisterra, Enzyme Microb. Technol. 24 (1999) 3–12.
[11] W. Piyatheerawong, Y. Iwasaki, X. Xu, T. Yamane, J. Mol. Catal. B:
Enzyme 28 (2004) 19–24.
[12] T. Nakaoki, Y. Mei, L.M. Miller, A. Kumar, B. Kalra, M.E. Miller, O. Kirk,
M. Christensen, R.A. Gross, Ind. Biotechnol. 1 (2005) 126–134.
[13] M. Petkar, A. Lali, P. Caimi, M. Daminati, J. Mol. Catal. B: Enzyme 39
(2006) 83–90.
[14] O. Orçaire, P. Buisson, A.C. Pierre, J. Mol. Catal. B: Enzyme 42 (2006)
106–113.
[15] R.M. Blanco, P. Terreros, N. Muñoz, E. Serra, J. Mol. Catal. B: Enzyme
47 (2007) 13–20.
[16] B. Chen, E.M. Miller, L. Miller, J.J. Maikner, R.A. Gross, Langmuir 23
(2007) 1381–1387.
[17] H.O. Finklea, in: R.A. Meyers (Ed.), Encyclopedia of Analytical Chemistry,
11, John Wiley & Sons, Chichester, 2000, pp. 10090–10115.
[18] K.J. Wiechelman, R.D. Braun, J.D. Fitzpatrick, Anal. Biochem. 175 (1988)
231–237.
[19] C.M. Stocscheck, Methods Enzymol. 182 (1990) 50–69.
[20] T. Vorderwülbecke, K. Kieslich, H. Erdmann, Enzyme Microb. Technol.
14 (1992) 631–639.
[21] F. Höök, B. Kasemo, T. Nylander, C. Fant, K. Scott, H. Elwing, Anal.
Chem. 73 (2001) 5796–5804.
[22] C. Larsson, M. Rodahl, F. Höök, Anal. Chem. 75 (2003) 5080–5087.
[23] J. Uppenberg, M.T. Hansen, S. Patkar, T.A. Jones, Structure 2 (1994)
293–308.
[24] K. Wannerberger, T. Arnebrant, Langmuir 13 (1997) 3488–3493.
[25] R.M. Blanco, P. Terreror, M. Fernández-Pérez, C. Otero, G. Dı́az-González,
J. Mol. Catal. B: Enzyme 30 (2004) 83–93.
[26] I. Petry, A. Ganesan, A. Pitt, B.D. Moore, P.J. Halling, Biotechnol. Bioeng.
95 (2006) 984–991.
[27] M.T. Reetz, R. Wenkel, D. Avnir, Synthesis (2000) 781–783.
J.A. Laszlo, K.O. Evans / Journal of Molecular Catalysis B: Enzymatic 48 (2007) 84–89
[28] A.O. Magnusson, J.C. Rotticci-Mulder, A. Santagostino, K. Hult, Chembiochem 6 (2005) 1051–1056.
[29] P. López-Serrano, L. Cao, F. van Rantwijk, R.A. Sheldon, Biotechnol. Lett.
24 (2002) 1379–1383.
89
[30] J.M. Palomo, M. Fuentes, G. Fernández-Lorente, C. Mateo, J.M. Guisan,
R. Fernández-Lafuente, Biomacromolecules 4 (2003) 1–6.
[31] S.H. Brewer, W.R. Glomm, M.C. Johnson, M.K. Knag, S. Franzen, Langmuir 21 (2005) 9303–9307.