Interactions of Endoglucanases with Amorphous Cellulose Films

3 May 2012 - ABSTRACT: A study of the interaction of four endoglucanases with amorphous cellulose films by neutron reflectometry (NR) and quartz...
0 downloads 0 Views 4MB Size
Download PDF
Article pubs.acs.org/Langmuir

Interactions of Endoglucanases with Amorphous Cellulose Films Resolved by Neutron Reflectometry and Quartz Crystal Microbalance with Dissipation Monitoring Gang Cheng,†,‡ Supratim Datta,†,¶ Zelin Liu,§ Chao Wang,§ Jaclyn K. Murton,‡ Page A. Brown,‡ Michael S. Jablin,∥ Manish Dubey,∥ Jaroslaw Majewski,∥ Candice E. Halbert,⊥ James F. Browning,⊥ Alan R. Esker,§ Brian J. Watson,# Haito Zhang,# Steven W. Hutcheson,# Dale L. Huber,‡,▽ Kenneth L. Sale,†,‡ Blake A. Simmons,†,‡ and Michael S. Kent*,†,‡ †

Joint BioEnergy Institute, Emeryville, California, United States Sandia National Laboratories, Livermore, California and Albuquerque, New Mexico, United States § Department of Chemistry, Virginia Tech, Blacksburg, Virginia, United States ∥ Lujan Neutron Science Center, Los Alamos National Laboratories, Los Alamos, New Mexico, United States ⊥ Spallation Neutron Source, Oak Ridge National Laboratory, Oak Ridge, Tennessee, United States # Department of Cell Biology and Molecular Genetics, University of Maryland, College Park, Maryland, United States ▽ Center for Integrated Nanotechnologies, Albuquerque, New Mexico, United States ‡

S Supporting Information *

ABSTRACT: A study of the interaction of four endoglucanases with amorphous cellulose films by neutron reflectometry (NR) and quartz crystal microbalance with dissipation monitoring (QCM-D) is reported. The endoglucanases include a mesophilic fungal endoglucanase (Cel45A from H. insolens), a processive endoglucanase from a marine bacterium (Cel5H from S. degradans), and two from thermophilic bacteria (Cel9A from A. acidocaldarius and Cel5A from T. maritima). The use of amorphous cellulose is motivated by the promise of ionic liquid pretreatment as a second generation technology that disrupts the native crystalline structure of cellulose. The endoglucanases displayed highly diverse behavior. Cel45A and Cel5H, which possess carbohydrate-binding modules (CBMs), penetrated and digested within the bulk of the films to a far greater extent than Cel9A and Cel5A, which lack CBMs. While both Cel45A and Cel5H were active within the bulk of the films, striking differences were observed. With Cel45A, substantial film expansion and interfacial broadening were observed, whereas for Cel5H the film thickness decreased with little interfacial broadening. These results are consistent with Cel45A digesting within the interior of cellulose chains as a classic endoglucanase, and Cel5H digesting predominantly at chain ends consistent with its designation as a processive endoglucanase.

■

INTRODUCTION Cellulases are enzymes that hydrolyze β-(1−4)-bonds in cellulose and have been widely used in many industrial applications.1 Recently, the significance of cellulose hydrolysis in the context of conversion of lignocellulosic biomass to fuels and other commodity chemicals has come to the forefront.2 Cellulose hydrolysis is a heterogeneous reaction in which cellulases initially bind to the surface of the insoluble cellulose substrate, followed by subsequent hydrolysis of glycosidic bonds.3,4 The efficient enzymatic hydrolysis of cellulose to glucose requires the synergistic action of three types of cellulases: endoglucanases, exoglucanases, and β-glucosidases. Many cellulases are composed of a catalytic domain to which a carbohydrate-binding module (CBM) is attached at either the N- or the C-terminus. Classic endoglucanases hydrolyze the β1,4-glycosidic bonds in the interior of the chains while exoglucanases (or cellobiohydrolases) hydrolyze from the © 2012 American Chemical Society

chain ends in a processive manner and release cellobiose. Processive endoglucanases are preferentially active near chain ends, and also release cellobiose. β-Glucosidases convert cellobiose units to glucose. Unraveling the interactions between different types of cellulases and the insoluble substrate is a prerequisite for the design of more effective enzyme systems.5 However, the heterogeneous nature of the supramolecular structure of lignocellulosic biomass and the synergistic action of different enzymes make it difficult to understand the complicated interactions that occur between enzyme cocktails and substrates.3−6 One approach to investigate cellulase binding and substrate degradation in a controlled way has been to use model films of Received: October 6, 2011 Revised: April 28, 2012 Published: May 3, 2012 8348

dx.doi.org/10.1021/la300955q | Langmuir 2012, 28, 8348−8358

Langmuir

Article

reported elsewhere.18 The A. acidocaldarius Cel9A clone was a gift from Prof. E. Schneider (Humboldt-Universität zu Berlin, Institut für Biologie/Bakterienphysiologie, Germany).19 Both enzymes were expressed in E. coli BL 21(DE3) (Novagen) and purified by nickel affinity chromatography using a 5 mL Hi-trap column (GE Life Sciences). The protein concentrations were measured by Bradford Assay for both recombinant enzymes. The recombinant protein purity was visually assessed using SDS-PAGE. The expression and purification of Cel5H was described previously.15 Cel45A (also called EG V) from Humicola insolens was obtained from Novozymes as the semicommercial product Novozyme 476. This enzyme product has been used in several prior studies.7,9,20 The enzyme is expressed in a recombinant fungal host and secreted as the predominant protein but is not purified. SDS-PAGE analysis shows a broad band from 40−60 kDa. The protein concentration was measured by Bradford Assay. Cel5A was dissolved in 50 mM sodium acetate buffer pH 5.0. Cel9A was dissolved in 50 mM sodium acetate buffer pH 7.0. Cel5H was dissolved in 50 mM sodium phosphate pH 6.5 with 300 mM NaCl. Cel45A was dissolved in 50 mM sodium acetate buffer pH 4.5 following the method of Josefsson et al.9 Preparation of Regenerated Cellulose Films. Smooth, uniform regenerated cellulose films were prepared on polished silicon wafers (diameter = 75 mm, thickness = 5 mm) and on QCM sensors (diameter = 14 mm and thickness = 0.3 mm) from precursor films of trimethylsilylcellulose (TMSC). TMSC was prepared from microcrystalline cellulose as reported elsewhere.21 For the QCM-D studies, the TMSC films were converted to cellulose by exposure to the vapors of a 3.3 N HCl solution for 3 min. During the course of this study, we discovered that the conversion conditions strongly impact the surface roughness of the films. In particular, the above-mentioned conversion conditions resulted in surface roughnesses that seriously limited the information that could be obtained from NR. For the NR study, the films were converted by exposure to the vapors of a 0.5 N HCl solution for 15 min. These conditions are sufficient to achieve full conversion of TMSC to cellulose and result in highly smooth surfaces.11 The thicknesses of the regenerated cellulose films on the QCM-D substrates were determined by angle resolved laser (632.8 nm, He−Ne laser) ellipsometry (Picometer Ellipsometer, Beaglehole Instruments, Wellington, New Zealand). The data were collected in 1.0° intervals from 65° to 80°. Measurements were made at three spots and reported as the average thickness plus-or-minus one standard deviation. The thicknesses of the regenerated cellulose films in air were 300 ± 10 Å. For the NR samples, the dry film thicknesses were determined from NR and ranged from 240 to 310 Å. This range of thicknesses resulted as conditions (solution concentration, spinning speed and acceleration, extent of repeated filtration, and filter size) were varied in an attempt to minimize the surface roughness. Neutron Reflectivity. NR measures the ratio of reflected to incident intensity as a function of momentum transfer qz = 4π sin θ/λ, where θ is the angle of incidence with respect to the plane of the film surface and λ is the wavelength.22 The form of this curve is determined by the in-plane average scattering length density (SLD also denoted b/ v) profile normal to the surface. The SLD is directly related to the atomic composition and density.22 NR studies were performed on the SPEAR reflectometer (Lujan Center/LANSCE) and the Liquids Reflectometer (SNS/ORNL). Both reflectometers operate in the timeof-flight mode where a band of wavelengths impinge onto the sample and are resolved at the detector on the basis of their time-of-flight. The measurements at 20 °C were performed as described in our prior report.11 To perform measurements at elevated temperature, the sample cell was placed between two copper blocks through which a heating fluid was circulated, and the sample cell and copper blocks were enclosed in a Styrofoam enclosure with thin aluminum foil windows. The temperature of the sample cell was continuously monitored with a thermocouple. Temperatures up to 75 °C were maintained within ±2 °C throughout the course of each measurement series. The volume fraction of cellulose was determined from the measured SLD of the swelled film and the SLD values of water (or buffer) and

regenerated cellulose in combination with ellipsometry and quartz crystal microbalance with dissipation monitoring (QCM-D).7−9 In QCM-D, the resonant frequency and dissipation of a quartz crystal sensor are measured. The frequency is sensitive to the total mass coupled with the oscillating crystal, which in the present case includes cellulose, adsorbed enzyme, and strongly associated water molecules. The important contribution of water within swollen films in QCMD measurements has been discussed previously.10 Energy dissipated during each oscillation is sensitive to changes in film viscoelasticity, either within the bulk of the film or at the surface.8,9 Neutron reflectometry (NR), introduced recently as a technique for analysis of the interactions of cellulases with cellulose, can resolve the volume fraction profile of water through cellulose films.11,12 This reveals whether an enzyme acts on the surface or throughout the bulk of the film, and whether its activity results in removal of mass, increased water swelling, or increased surface roughness. Thus, QCM-D and NR are highly complementary and can provide a wealth of detailed information about the interactions of cellulases with cellulose films. A combined NR and QCM-D study of the interaction of a fungal enzyme extract (T. viride) and an endoglucanase from A. niger (CelAN) with amorphous cellulose films demonstrated the complementarity of these two techniques.11 It was found that at 20 °C the endoglucanase, which lacks a CBM, released mass mainly from the surface of the film and did not increase the surface roughness. By contrast, over the same time period, the T. viride extract rapidly digested the entire film, initially roughening the film surface followed by penetration and activity throughout the bulk of the film, demonstrating the synergistic effect of a cocktail of cellulases. In this work, NR and QCM-D results were compared for four endoglucanases: Cel45A, Cel5H, Cel9A, and Cel5A. The endoglucanases were examined separately to reveal the mode of interaction between each enzyme and the cellulose substrate. Cel45A from H. insolens, with a total molecular weight of 43 kDa, consists of a catalytic domain of 22.9 kDa and a binding module separated by a flexible linker of 33 amino acids.13 The binding module belongs to the CBM1 family, the same classification family as Cel7A of T. reesei.14 Cel45A has an optimum temperature of 50 °C. Cel5H from the marine bacterium S. degradans is a highly active processive endoglucanase with total molecular weight of 68 kDa and an optimum temperature of 50 °C.15 It consists of a GH5 catalytic module and a family 6 CBM. The endoglucanase Cel9A from A. acidocaldarius belongs to glycoside hydrolase family 9 (GH9) and consists of an N-terminal immunoglobulin-like (Ig) domain followed by the catalytic domain with a molecular weight of 59 kDa and an optimum temperature of 70 °C.16 Cel5A from T. maritima with molecular weight 40 kDa belongs to the GH5 family and contains only a catalytic domain. It has an optimum temperature of 80 °C.17 As reported below, the combined NR and QCM-D measurements reveal highly diverse behavior for the four endoglucanases.

■

MATERIALS AND EXPERIMENTS

Sodium hydroxide, acetic acid, microcrystalline cellulose (Avicel PH101), hexamethyldisilazane, dimethylacetamide, lithium chloride, tetrahydrofuran, and methanol were all purchased from Sigma-Aldrich. Cel5A from Thermotoga maritima (NP_229549) was cloned by a ligation independent cloning method into the expression vector, pCDF2 LIC/Ek (Novagen/EMD Chemicals, Gibbstown, NJ) as 8349

dx.doi.org/10.1021/la300955q | Langmuir 2012, 28, 8348−8358

Langmuir

Article

Figure 1. (a) NR data for a regenerated cellulose film exposed to a solution of Cel45A at 5 μM and 20 °C. (b) Cellulose volume fraction profiles obtained from the fitting analysis.

■

amorphous cellulose using the following relation (which assumes additivity of volumes):

(b/v)meas = ϕcellulose(b/v)cellulose + (1 − ϕcellulose)(b/v)water

RESULTS The information derived from NR and QCM-D will be reported for each individual enzyme. We note that while interpreting the NR data in terms of the volume fraction profiles of water and cellulose is straightforward, interpreting the QCM-D data is more challenging. In particular, it is important to note that due to the compensating effects on Δf of cellulose mass lost, enzyme adsorption, and changes in the mass of water coupled to the film, the magnitude and sign of Δf may not vary in a simple manner with the level of enzyme activity. In addition, the effect on Δf will be different for cellulose mass loss from the bulk of the film compared with cellulose mass lost from the surface of the film, since cellulose mass lost from the bulk will be partially compensated by increased water content within the film. Quantifying the amount of coupled water from the NR-derived volume fraction profiles is not straightforward, as the entire volume of water within the film may not be mechanically coupled to the film. Mechanically coupled water includes hydration layers that are tightly bound to hydroxyl groups of cellulose chains but may also include additional layers that extend further from the chains. To our knowledge, this is not well understood. Cel45A. NR scans upon injecting Cel45A at 5 μM and 20 °C are shown in Figure 1a. Corresponding volume fraction profiles are given in Figure 1b. Upon introducing Cel45A, the fringes in the NR curve shifted to lower qz and the magnitude of the fringes decreased sharply with greater dampening at higher qz. The former effect indicates that the film thickness increased, in contrast with the results for CelAN, an endoglucanase from A. niger lacking a CBM reported previously where digestion occurred mainly at the surface.11 The dampening of the fringes over most of the qz range for the scan initiated only 15 min after introducing the enzyme also contrasts sharply with the prior results for CelAN. This indicates that the activity of Cel45A resulted in a large broadening of the solution−film interface. With longer exposure time, the curves shifted upward over most of the qz range, indicating substantial loss of cellulose mass from the film. The volume fraction profiles show that the film thickness

(1)

where ϕcellulose is the volume fraction of cellulose, (b/v)cellulose = 1.67 × 10−6 Å−2 is the SLD of the cellulose film,11 and (b/v)water = −0.54 × 10−6 Å−2 is the SLD of water. The volume fraction profiles are represented by color−coded bands that indicate uncertainty limits determined from a Monte Carlo resampling procedure.23 QCM-D Measurements. A Q-Sense E4 quartz crystal microbalance was used to investigate the adsorption and activity of the enzymes on regenerated cellulose films. The measurement procedure was described in a previous report.11 Frequency (Δf) and dissipation (ΔD) changes for the fundamental frequency (4.95 MHz for gold coated quartz crystals) and six odd overtones (n = 3−13) were monitored simultaneously. Normally, Δf and ΔD from the first overtone were noisy because of insufficient energy trapping and were not included in the graphs. If the adsorbed mass is evenly distributed, rigidly attached, and small compared to the mass of the crystal, Δm, mass per unit area, can be calculated by the Sauerbrey equation:9 Δm = −

C Δf n

(2)

where n is the overtone number and C is a constant (0.177 mg·m−2·Hz−1). However, the linear relationship between the adsorbed mass and the frequency change is not valid for viscoelastic layers adsorbed onto solid surfaces as in this study. Moreover, the NR data show that the films can become highly inhomogeneous as enzymatic digestion proceeds. So qualitative comparisons of Δf for different enzymes are given rather than attempting to estimate quantitative changes in mass. By measuring the dissipation of energy in the adsorbed layers simultaneously with the frequency change, information is obtained about the stiffness of the adsorbed layer. The dissipation factor is defined as the ratio between the energy dissipated and the energy stored during a single oscillation:9

D=

Edissipated 2πEstored

(3)

Variation of scaled frequency, Δf/n, and D with overtone number is indicative of film viscoelasticity. 8350

dx.doi.org/10.1021/la300955q | Langmuir 2012, 28, 8348−8358

Langmuir

Article

the film thickness increased but to a lesser degree than at 20 °C. Also, more mass was released from the film (Table 1) and a greater uptake of water occurred than at 20 °C, as expected at Topt. For the full scan initiated 6 h after injecting the enzyme shown in Figure 2a, the water volume fraction was 0.82 compared with 0.56 in the as-prepared film. QCM-D data for Cel45A adsorbed onto a regenerated cellulose film at 20 °C are shown in Figure 3a. Results are shown for the third, fifth, seventh, ninth, and eleventh overtone. Δf dropped and a large increase in dissipation occurred shortly after introducing Cel45A (within 20 min). The initial decrease in frequency and increase in dissipation are greater than observed previously with CelAN. Both Δf/n and ΔD were strongly dependent on overtone, which also contrasts with the prior results for CelAN. The initial rapid drop in frequency indicates a net increase in total mass (cellulose + enzyme + water) coupled with the oscillating crystal. These initial changes in Δf/n and ΔD are likely due to enzyme adsorption and perhaps also increased water swelling within the bulk of the film. The fact that the frequency was nearly constant after 1 h indicates the total mass coupled to the oscillating crystal remained nearly constant. Since NR shows mass loss over at least several hours, we conclude that after 1 h the cellulose mass loss was nearly compensated by increased water uptake. QCM-D data for Cel45A adsorbed onto a regenerated cellulose film at 50 °C are shown in Figure 3b. Following a brief period (several minutes) of enzyme adsorption, the frequency began to increase sharply, in contrast to the results at 20 °C, indicating a net mass loss. The increase in dissipation was weaker at 50 °C than at 20 °C. Both Δf/n and ΔD were again dependent on overtone but not as strongly as at 20 °C. Summary for Cel45A. The NR results show that Cel45A rapidly penetrates into and digests within the bulk of the cellulose films. Evidence for this has been reported previously.7,9 The data show that the activity of Cel45A caused the film thickness to increase substantially, more so at 20 °C than at 50 °C. While swelling and expansion of amorphous cellulose has been predicted for classic endoglucanases,9 to our knowledge this is the first direct observation of this effect. We

increased ∼40 Å for the scan initiated 15 min after injecting the enzyme, and the water volume fraction within the bulk of the film increased from 0.57 to 0.63. With even longer exposure time, the film thickness increased slightly but the water content in the bulk of the film increased substantially. For the scan initiated 2 h 15 min after injecting Cel45A, the water volume fraction within the bulk of the film was 0.77. The percentage of mass lost from the film in each case, determined by integrating the volume fraction profiles, is given in Table 1. Table 1. Percent Cellulose Mass Loss Determined from the NR Profiles enzyme

conc

Cel45A

5 μM

20 °C

temp

Cel5H

3.7 μM

50 °C 20 °C

50 °C

a

Cel9A

5 μM

Cel5A

5 μM

20 20 20 65 20 20 80

°C °C, 30 m 65 °C °C, 90 m 65 °C °C °C C, 30 m 80 °C °C

time

% mass loss

0.25 h 2.25 h 1.25 h 3.7 h 5.0 h 21 h 0.25 h 2.25 h 5.25 h 14.25 h

7.7 38 54 15 19 24 24 47 63 72 5.2 15 21 11 1.7 11 4.3

a

a a

a

Data collected after scans became independent of time.

NR scans for Cel45A at 5.0 μM and 50 °C (Topt) are shown in Figure 2a. The corresponding volume fraction profiles are given in Figure 2b. A series of 1 h scans over a limited qz range were collected upon injecting the enzyme. These scans (Figure S1, Supporting Information) show only small changes after the first scan initiated 15 min after injecting the enzyme. At 50 °C,

Figure 2. (a) NR data for a regenerated cellulose film exposed to a solution of Cel45A at 5 μM and 50 °C. (b) Cellulose volume fraction profiles obtained from the fitting analysis. 8351

dx.doi.org/10.1021/la300955q | Langmuir 2012, 28, 8348−8358

Langmuir

Article

Figure 3. Δf/n and ΔD versus time from QCM-D for regenerated cellulose films exposed to 5 μM Cel45A (a) at 20 °C and (b) at 50 °C. Arrows indicate where solutions were switched.

Figure 4. (a) NR data for a regenerated cellulose film exposed to a solution of Cel5H at 3.7 μM and 20 °C. (b) Cellulose volume fraction profiles obtained from the fitting analysis. The red and purple bands correspond to scans initiated 3.7 and 21 h after injecting the enzyme, respectively. Those scans are shown in Figure S2 in the Supporting Information.

believe that the large increase in roughness of the film−solution interface is a consequence of film expansion. Film expansion is likely due to the endoglucanase digesting within the interior of chains, substantially reducing the average molecular weight. Among the enzymes studied in our present and prior work,11 substantial film expansion occurred only for Cel45A. The fact that Δf/n and ΔD varied with overtone indicates increased viscoelasticity.

At 50 °C, the results were in several respects noticeably different than at 20 °C. A large interfacial broadening again occurred upon enzyme injection. However, in contrast to the results at 20 °C, a weaker film expansion was observed at 50 °C by NR, and Δf began to increase sharply within minutes following the initial drop, in contrast to the very weak increase observed at 20 °C. Finally, the increase in dissipation was also weaker at 50 °C than at 20 °C. These three results are consistent with greater digestion at the surface relative to the 8352

dx.doi.org/10.1021/la300955q | Langmuir 2012, 28, 8348−8358

Langmuir

Article

Figure 5. (a) NR data for a regenerated cellulose film exposed to a solution of Cel5H at 3.7 μM and 50 °C. (b) Cellulose volume fraction profiles obtained from the fitting analysis. The purple and dark blue profiles correspond to scans initiated 8.25 and 12.25 h, respectively, after injecting the enzyme.

Figure 6. Δf/n and ΔD versus time from QCM-D for regenerated cellulose films exposed to 3.7 μM Cel5H at (a) 20 °C and (b) 50 °C. Arrows indicate where solutions were switched.

We note that the initial drop in Δf was weaker at 50 °C than at 20 °C, which we attribute to stronger adsorption at lower temperature as expected from entropic considerations. Cel5H. NR scans for Cel5H at 3.7 μM and 20 °C are shown in Figure 4a. NR data collected 5 h after injecting Cel5H are compared with data prior to adding the enzyme. Scans collected 3.7 and 21 h after injecting Cel5H are shown in Figure S2 in the Supporting Information. Volume fraction profiles from all scans are given in Figure 4b. The changes in the NR data with time for Cel5H are qualitatively different than

bulk of the film at 50 °C than at 20 °C. Whereas film expansion and increased water swelling occur due to cleavage of cellulose chains within the bulk of the film, high activity localized at the surface of the film will result in mass loss from the surface and a decrease in film thickness. Digestion at the surface will result in a much greater increase in Δf due to the loss of cellulose mass and the associated coupled water. Finally, digestion at the surface will result in a weaker increase in dissipation than digestion in the bulk, since for Cel45A the latter will result in a large increase in viscoelasticity as a consequence of swelling. 8353

dx.doi.org/10.1021/la300955q | Langmuir 2012, 28, 8348−8358

Langmuir

Article

Figure 7. (a) NR data for a regenerated cellulose film exposed to a solution of Cel9A at 5 μM and 20 °C. Data are also shown after subsequent heating at 65 °C for 30 and 90 min. (b) Cellulose volume fraction profiles obtained from the fitting analysis. The colors of the profile bands correspond to the colors of the data in (a).

adsorption, followed quickly by a strong increase. The increase was stronger than that reported previously for CelAN.11 The dissipation increased upon adsorption and continued to increase for the majority of the run. At 50 °C (Figure 6b), the initial drop in Δf was weaker than at 20 °C, indicating weaker adsorption. Following the adsorption phase, Δf again increased, but the increase in Δf was substantially weaker at 50 °C than at 20 °C. Also, the dissipation went through a maximum, in contrast to the steady increase following the adsorption phase at 20 °C. Summary for Cel5H. In contrast to the results for Cel45A, NR showed that the film did not expand with activity of Cel5H but rather the film thickness decreased. Also in contrast to the results for Cel45A, with Cel5H the film−solution interface broadened only slightly. Regarding the changes in Δf following the rapid adsorption stage, a temperature dependence opposite to that for Cel45A was observed. At 50 °C the increase in Δf was weaker than at 20 °C, despite the fact that Cel5H is known to be more active at 50 °C,15 a fact confirmed by the NR data. This suggests that the ratio of the increase in mass of coupled water to cellulose mass lost is greater at 50 °C than at 20 °C, opposite to the trend for Cel45A. The fact that at 50 °C the dissipation went through a maximum is reminiscent of the results for digestion of cellulose films with fungal extracts reported previously.8,11 In those cases, a maximum occurred as the mass within the film became substantially depleted. The same explanation may apply here, as at 50 °C the water volume fraction within the film increased greatly reaching a final value of 0.86. Cel9A. NR data upon injecting Cel9A at 5 μM at 20 °C are shown in Figure 7a, and the corresponding volume fraction profiles are given in Figure 7b. During the first 30 min, the fringes in the NR curve shifted to slightly higher qz with a small increase in the fringe spacing, indicating a decrease in film thickness. The volume fraction profiles contrast with those for Cel5H and Cel45A at 20 °C in that the changes are much weaker and most of the digestion occurs at the surface of the

those observed for Cel45A. Upon addition of Cel5H, the fringes shift to higher qz with increased spacing between successive fringes, indicating a decrease in film thickness. The magnitude of the fringes was slightly damped over the entire qz range, indicating increased water content within the film. Specifically, the volume fraction profiles show that the film thickness decreased ∼20 Å within the first 3.7 h and the water volume fraction within the film increased from 0.55 to 0.60. With further incubation, the positions of the fringes remained nearly constant while decreasing in amplitude and the reflectivity increased over most of the qz range (Figure S2, Supporting Information). These changes indicate that the film thickness stayed nearly constant while the water content within the film continued to slowly increase, reaching a volume fraction of 0.64 after 21 h. The fact that the fringes persisted for all scans indicates that the roughness of the film−solution interface increased only slightly during the entire digestion period, in contrast to the results for Cel45A. NR data for Cel5H at 3.7 μM and 50 °C (Topt) are shown in Figure 5. In Figure 5a, a scan for the as-prepared film in buffer is compared with scans initiated 0.25 and 2.25 h after injecting the enzyme. Profiles corresponding to these scans as well as scans initiated 8.25 and 12.25 h after injecting Cel5H are shown in Figure 5b. Much greater changes were observed at Topt than at 20 °C. The characteristics of the profiles are qualitatively similar to those measured at 20 °C, but the changes occurred more rapidly and to a greater extent. For the scan initiated 15 min after injecting Cel5H, the volume fraction profile shows the film thickness decreased by 45 Å and the volume fraction of water in the bulk of the film increased from 0.54 to 0.66. These trends continued for many hours but at a progressively slower rate. After 14 h, the film thickness was 80 Å less than the asprepared film, and the water volume fraction had increased to 0.86. QCM-D data for Cel5H adsorbed onto a regenerated cellulose film at 20 °C are shown in Figure 6a. Δf dropped substantially upon introducing Cel5H due to enzyme 8354

dx.doi.org/10.1021/la300955q | Langmuir 2012, 28, 8348−8358

Langmuir

Article

Figure 8. (a) NR data for a regenerated cellulose film exposed to a solution of Cel9A at 5 μM and 65 °C. Cellulose volume fraction profiles from the fitting analysis are shown in (b).

Figure 9. Δf/n and ΔD versus time from QCM-D for regenerated cellulose films exposed to 5 μM Cel9A at (a) 20 °C and (b) 50 °C. Arrows indicate where solutions were switched.

65 °C resulted in only a very slight additional decrease in film thickness and increase in water content. NR data were also collected upon injecting Cel9A at 5 μM at 65 °C, close to Topt for soluble substrates. Scans of 1 h duration were collected over a limited range of qz for 8 h. Little change was observed after the second scan (Supporting Information, Figure S3). An NR scan over the full qz range was collected after 8 h, and the data are compared with the data for the asprepared film in Figure 8a. The corresponding volume fraction profiles are shown in Figure 8b. In contrast to the NR results for Cel5H and Cel45A, at Topt the NR data show relatively small effects, consisting of a small shift to higher qz and a small decrease in the magnitude of the fringes. However, in this case,

film. Fitting analysis indicated a decrease in thickness of 40 Å. The magnitude of the fringes became progressively damped at higher qz values, indicating a slight broadening of the interface between film and solution. After 30 min, no further changes were detected, and this result was consistent over several 30 min scans. The sample was then placed into an oven at 65 °C for 30 min. The sample was then removed from the oven, and a NR scan was collected at 20 °C. The film decreased by another 60 Å, and the interface broadened still further. In addition, a small increase in water content within the bulk of the film was detected (volume fraction of 0.67 compared with 0.64 for the as-prepared film). Further heating for an additional 60 min at 8355

dx.doi.org/10.1021/la300955q | Langmuir 2012, 28, 8348−8358

Langmuir

■

the fringes are not progressively damped at higher qz. These changes correspond to a decrease in film thickness of 15 Å and a slight increase in water content within the bulk of the film from 0.48 to 0.54. The percent mass lost from the film in this case was about half of that in Figure 7b, where adsorption took place at 20 °C and the system was subsequently heated to Topt. QCM-D data for Cel9A adsorbed from sodium acetate buffer onto regenerated cellulose films at 20 °C are shown in Figure 9a. The changes in Δf and ΔD were much weaker than for Cel45A and Cel5H. The value of Δf dropped quickly, then increased rapidly, then increased more gradually to a plateau prior to flushing with buffer. When pure buffer was flowed through the cell following several hours of incubation, the frequency returned to the initial baseline. The variation with overtone number is modest for Cel9A. The small variation with overtone number may be due to the presence of a layer of adsorbed enzyme. ΔD increased abruptly and then plateaued upon injecting Cel9A. The plateau value was small compared with the increases observed upon addition of Cel45A and Cel5H. Upon rinsing with buffer, ΔD decreased back to nearly its initial value, consistent with removal of adsorbed enzyme. QCM-D studies performed with Cel9A at 50 °C are shown in Figure 9b. The changes were very weak and close to the limit of detectability. The results show a much smaller drop in Δf upon introducing the enzymes than observed at 20 °C. This is consistent with weaker adsorption at 50 °C compared with that at 20 °C. Consistent with this conclusion, the variation with overtone number was also weaker at 50 °C than at 20 °C. Δf gradually decreased with time following the initial period of enzyme adsorption, in contrast to the results at 20 °C, where Δf increased with time following enzyme adsorption. Cel5A. NR and QCM-D data for Cel5A are shown in Figure S4−S6 (Supporting Information). The changes following enzyme addition are even weaker than observed for Cel9A. Summary for Cel9A and Cel5A. A comparison of the initial drop in Δf indicates greater adsorption of Cel9A onto the cellulose surface than Cel5A at 20 °C. The NR measurements for Cel9A at 20 °C show loss of ∼40 Å of material from the outer surface of the film during the first 30 min (Figure 7). This is consistent with the observed small increase in Δf from QCMD after the initial sharp drop during the adsorption phase (Figure 9a). The dissipation remained constant following the initial adsorption period, indicating that no detectable change occurred in the film stiffness. This is consistent with enzyme activity primarily at the outer surface of the film. After flowing pure buffer through the QCM-D cell, Δf and ΔD returned to values that were close to the original values. The increase in Δf and decrease in ΔD upon flowing buffer through the cell is most likely due to removal of adsorbed enzyme. The return of Δf to nearly its original value suggests that the amount of Cel9A remaining on the substrate after rinsing compensates for the small loss of cellulose from the film. At elevated temperature, the changes observed by both techniques were very weak. The NR data at 65 °C show that Cel9A activity led to a modest increase in water content within the film, indicating some enzyme digestion within the bulk. Together the NR and QCM-D data show that for Cel9A at Topt there is weaker adsorption and less mass lost from the surface of the film, but greater penetration and digestion within the bulk of the film than at 20 °C. However, the extent of digestion within the bulk is very modest compared with the results for Cel5H and Cel45A. The results for Cel5A indicated even weaker digestion than for Cel9A both at 20 °C and at Topt.

Article

DISCUSSION

As expected, the amount of enzyme adsorption was greater for the two enzymes with a CBM (Cel45A and Cel5H) than for the enzymes lacking a CBM (Cel9A, Cel5A, and CelAN reported previously11). Also, enzyme adsorption was consistently weaker at 50 °C than at 20 °C, as expected based on entropic considerations. The NR measurements revealed much greater activity within the bulk of the cellulose films for the two endoglucanases (Cel45A and Cel5H) that have CBMs compared with those lacking CBMs (Cel9A and Cel5A). For the five endoglucanases examined in our present and prior work there is a correlation between the extent of penetration into the film and the presence of a CBM. Rapid penetration and substantial digestion within the bulk of the film was observed for Cel45A and Cel5H, which contain CBMs, whereas far weaker penetration into the film was observed for Cel9A, Cel5A, and CelAN which lack CBMs. At 20 °C more substantial digestion within the bulk of the film was observed with Cel45A, which has a family 1 CBM, than for Cel5H, which has a family 6 CBM. The ability of Cel45A to rapidly penetrate into amorphous cellulose films is somewhat surprising as the CBM1 family is known to bind to crystalline cellulose through three aromatic residues that form a flat surface.24−26 Truncated forms of both Cel5H and Cel45A lacking the CBMs have been reported to have substantially reduced activity on insoluble cellulose,7,9,15 demonstrating the critical importance of the CBMs. Despite the fact that both Cel5H and Cel45A penetrate and digest within the bulk of the films, very distinct differences were observed. The effects of digestion by Cel45A follow the expectations for a classic endoglucanase, as cleavage of internal bonds within the cellulose chains will lower the average molecular weight and increase the number of chain ends. For a glassy polymer, an increase in chain ends will result in increased free volume, a decrease in the glass transition temperature, and increased swelling.27 Substantial interfacial broadening will result as a consequence of increased swelling. On the other hand, Cel5H is a processive endoglucanase, a designation based on the observed release of cellobiose and high ratio of soluble to insoluble products.15 If Cel5H digests preferentially at chain ends, activity will not increase the number of chain ends and therefore will result in little increased swelling. Furthermore, digestion preferentially near chain ends will result in more mass lost in the form of soluble cellulose fragments compared to endoglucanases that cleave randomly within cellulose chains. Because of this, activity at the surface will result in a more substantial decrease in thickness than for classic endoglucanases. The different temperature dependence observed in the NR and QCM-D data for Cel45A and Cel5H suggests that enzyme penetration is driven by different mechanisms in the two cases. For Cel45A, the difference in the response of Δf at the two temperatures suggests that the relative amount of activity at the surface compared to the bulk is greater at Topt than at 20 °C. We hypothesize that this is due to changes in the relative rates of penetration and catalytic activity. We propose that the enzyme penetrates by the CBM hopping from one cellulose binding site to another, and that there is a competition between the rate of hopping and the rate of digestion. Due to the lower activity of the catalytic domain and stronger interaction of the CBM with cellulose at 20 °C, we suggest that the rate of enzyme penetration is high relative to the rate of bond cleavage. 8356

dx.doi.org/10.1021/la300955q | Langmuir 2012, 28, 8348−8358

Langmuir

Article

On the other hand, at Topt, we expect that the higher activity of the catalytic domain and weaker binding affinity of the CBM with cellulose results in more digestion near the surface of the film prior to enzyme penetration deep into the film. Since this enzyme cleaves internally within cellulose chains, a low level of activity of the catalytic domain is still sufficient to cause local swelling and enable the enzyme to penetrate further into the film. We refer to this as the “penetration by local swelling” model. On the other hand for Cel5H, the difference in the response of Δf at the two temperatures suggests that the relative amount of activity at the surface compared to the bulk is greater at 20 °C than at Topt. This is borne out in the measured profiles, as the ratio of mass loss from the surface versus the bulk is substantially greater in Figure 4b than in Figure 5b for the earliest digestion times. This, along with the striking difference in interfacial roughness with Cel5H compared with Cel45A, implies an entirely different mechanism of penetration. Since this enzyme digests primarily near chain ends, little increase in the number of chain ends and consequently little local swelling occurs. Instead, we suggest that the enzyme must burrow its way into the film by virtue of the activity of the catalytic domain. In this “penetration by burrowing” model, the greater activity of the catalytic domain at Topt causes this enzyme to penetrate more rapidly into the film. It is also possible that a difference in the binding properties of the two CBMs could play a role in the mechanism of penetration. In a previously reported QCM-D study of Cel45A on cellulose films, a monotonic decrease in frequency was observed at room temperature following the initial drop due to enzyme adsorption. This monotonic decrease in frequency was interpreted as due to an increase in bound (hydrated) water as digestion proceeded and that this endoglucanase did not release mass from the film.9 In addition to the gradual decrease in frequency, the dissipation increased substantially over many hours and was interpreted as indicating chain scission and increased viscoelasticity. The present results may be in qualitative agreement with these findings if one considers that the increase in D and decrease in f occur on a more rapid timescale in the present study. However, their conclusion that no mass was released from the cellulose film was in conflict with a prior ellipsometry study involving the same enzyme.7 In the ellipsometry study, small increases in mass were initially detected due to enzyme adsorption that were followed by much larger decreases in film mass as digestion proceeded. The present NR results resolve the conflicting conclusions of these QCM-D and ellipsometry studies. The volume fraction profile in Figure 1b shows both an increase in film thickness as well as a loss of cellulose mass from the film. Due to the increase in film thickness the increase in bound water will compensate for the loss of cellulose mass, explaining the nearly constant frequency following adsorption observed in the present QCMD study. The above discussion is only qualitative. A quantitative comparison is not warranted, as the film preparation methods were not identical in the three studies. The prior QCM-D study involved cellulose films prepared by dissolving pulp in Nmethylmorpholine N-oxide (NMMO). This procedure results in films that are partly crystalline (cellulose II).28 The ellipsometry study involved cellulose films prepared by dissolving microcrystalline cellulose in dimethylacetamide and lithium chloride (DMAC/LiCl). No crystallinity has been detected for films spun from DMAC/LiCl or for the TMSCderived cellulose films in the present work.11,12,28

The activities of the endoglucanases from thermophilic bacteria were very limited and showed modest dependence on temperature. At 20 °C, Cel9A and Cel5A resulted in only small effects at the surface of the films. The extent of digestion was far weaker and ceased more rapidly than for the fungal endoglucanase CelAN reported previously.11 Subsequent elevation of the temperature to Topt resulted in only a very limited increase in film degradation. The relatively limited activity may be due to the lack of affinity of the enzymes for the substrate, or else the buildup of a layer of nonproductively bound enzyme at the surface of the films. Both Cel9A and Cel5A showed greater release of mass when adsorption occurred at 20 °C and the temperature was subsequently increased to Topt, than when the enzymes were introduced at Topt. This can be explained by greater binding at lower temperature, expected based on entropic considerations and consistent with the QCM-D data.

■

SUMMARY QCM-D and NR are highly complementary and provide unprecedented insight into the effect of endoglucanases on the structure of cellulose films. QCM-D provides changes in total mass coupled to the oscillating substrate and in film stiffness, while NR reveals the profile of water through the film at nanometer resolution. These techniques have resolved very diverse behavior for five endoglucanases. A trend has emerged in that the enzymes possessing a CBM penetrate into and digest within the bulk of the films to a much greater extent than those lacking a CBM. While Cel5H and Cel45A each contain CBMs and were both highly active within the bulk of the film, striking differences in behavior were observed. For Cel45A, film expansion and a large interfacial broadening were observed, whereas with Cel5H film thickness decreased with little interfacial broadening. Moreover, an opposite temperature dependence for Cel5H and Cel45A suggests that penetration is driven by different mechanisms in each case. These differences appear to result because Cel45A digests in the interior of the cellulose chains as a classic endoglucanase,9 whereas Cel5H preferentially digests at the ends of the chains consistent with its designation as a processive endoglucanase.15

■

ASSOCIATED CONTENT

S Supporting Information *

NR scans for 5 μM Cel45A initiated 1 and 5 h after injecting the enzyme at 50 °C compared with data for the as-prepared film. NR scans for 3.7 μM Cel5H at 20 °C initiated 3.7 and 21 h after injecting the enzyme. Successive 1 h NR scans for 5 μM Cel9A at 65 °C. NR data for Cel5A at 20 °C and at 80 °C. QCM-D data for Cel5A at 5 μM at 20 °C and at 50 °C. This material is available free of charge via the Internet at http:// pubs.acs.org.

■

AUTHOR INFORMATION

Present Address ¶

Department of Biology, Indian Institute of Science Education and Research, Kolkata, India.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was part of the DOE Joint BioEnergy Institute (http://www.jbei.org) supported by the U.S. Department of 8357

dx.doi.org/10.1021/la300955q | Langmuir 2012, 28, 8348−8358

Langmuir

Article

(14) Voutilainen, S. P.; Puranen, T.; Siika-aho, M.; Lappalainen, A.; Alapuranen, M.; Kallio, J.; Hooman, S.; Viikri, L.; Vehmaanpera, J.; Koivula, A. Cloning, expression, and characterization of novel thermostable family 7 cellobiohydrolases. Biotechnol. Bioeng. 2008, 101, 515−528. (15) Watson, B. J.; Zhang, H. T.; Longmire, A. G.; Moon, Y. H.; Hutcheson, S. W. Processive endoglucanases mediate degradation of cellulose by Saccharophagus degradans. J. Bacteriol. 2009, 191, 5697− 5705. (16) Pereira, J. H.; Sapra, R.; Volponi, J. V.; Kozina, C. L.; Simmons, B.; Adams, P. D. Structure of endoglucanase Cel9A from the thermoacidophilic Alicyclobacillus acidocaldarius. Acta Crystallogr. D 2009, 65, 744−750. (17) Pereira, J. H.; Chen, Z. W.; McAndrew, R. P.; Sapra, R.; Chhabra, S. R.; Sale, K. L.; Simmons, B. A.; Adams, P. D. Biochemical characterization and crystal structure of endoglucanase Cel5A from the hyperthermophilic Thermotoga maritima. J. Struct. Biol. 2010, 172, 372−379. (18) Datta, S.; Holmes, B.; Park, J. I.; Chen, Z. W.; Dibble, D. C.; Hadi, M.; Blanch, H. W.; Simmons, B. A.; Sapra, R. Ionic liquid tolerant hyperthermophilic cellulases for biomass pretreatment and hydrolysis. Green Chem. 2010, 12, 338−345. (19) Eckert, K.; Zielinski, F.; Lo Leggio, L.; Schneider, E. Gene cloning, sequencing, and characterization of a family 9 endoglucanase (CeIA) with an unusual pattern of activity from the thermoacidophile Alicyclobacillus acidocaldarius ATCC27009. Appl. Microbiol. Biotechnol. 2002, 60, 428−436. (20) Eriksson, J.; Malmsten, M.; Tiberg, F.; Callisen, T. H.; Damhus, T.; Johansen, K. S. Enzymatic degradation of model cellullose films. J. Colloid Interface Sci. 2005, 284, 99−106. (21) Kontturi, E.; Thune, P. C.; Niemantsverdriet, J. W. Cellulose model surfaces-simplified preparation by spin coating and characterization by X-ray photoelectron spectroscopy, infrared spectroscopy, and atomic force microscopy. Langmuir 2003, 19, 5735−5741. (22) Penfold, J.; Thomas, R. K. The application of the specular reflection of neutrons to the study of surfaces and interfaces. J. Phys.: Condens. Matter 1990, 2, 1369−1412. (23) Heinrich, F.; Ng, T.; Vanderah, D. J.; Shekhar, P.; Mihailescu, M.; Nanda, H.; Losche, M. A new lipid anchor for sparsely tethered bilayer lipid membranes. Langmuir 2009, 25, 4219−4229. (24) Beckham, G. T.; Matthews, J. F.; Bomble, Y. J.; Bu, L.; Adney, W. S.; Himmel, M. E.; Nimlos, M. R.; Crowley, M. F. Identification of amino acids responsible for processivity in a family 1 carbohydratebinding module from a fungal cellulase. J. Phys. Chem. B 2010, 114, 1447−1453. (25) Linder, M.; Mattinen, M.-L.; Kontteli, M.; Lindeberg, G.; Stahlberg, J.; Drakenberg, T.; Reinikainen, T.; Pettersson, G.; Annila, A. Identification of functionally important amino acids in the cellulosebinding domain of Trichoderma reesei cellobiohydrolase I. Protein Sci. 1995, 4, 1056−1064. (26) Takashima, S.; Ohno, M.; Hidaka, M.; Nakamura, A.; Masaki, H.; Uozumi, T. Correlation between cellulose binding and activity of cellulose binding domain mutants of Humicola grisea cellobiohydrolase I. FEBS Lett. 2007, 581, 5891−5896. (27) Hiemenz, P. Polymer Chemistry; Marcel Dekker, Inc.: New York, 1984. (28) Aulin, C.; Ahola, S.; Josefsson, P.; Nishino, T.; Hirose, Y.; Osterberg, M.; Wagberg, L. Nanoscale cellulose films with different crystallinities and microstructutres - their surface properties and interaction with water. Langmuir 2009, 25, 7675−7685.

Energy, Office of Science, Office of Biological and Environmental Research, through contract DE-AC02-05CH11231 between Lawrence Berkeley National Laboratory and the U.S. Department of Energy. The research at the Los Alamos Neutron Science Center using SPEAR reflectometer and the Spallation Neutron Source at Oak Ridge National Laboratory was sponsored by the Scientific User Facilities Division, Office of Basic Energy Sciences, U.S. Department of Energy. Los Alamos Neutron Science Center (LANSCE) is supported by DOE contract W7405-ENG-36. A.R.E., Z.L., and C.W. were supported as part of the Center for LignoCellulose Structure and Formation, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Award Number DE-SC0001090. This work was performed, in part, at the Center for Integrated Nanotechnologies, a U.S. Department of Energy, Office of Basic Energy Sciences user facility at Los Alamos National Laboratory (Contract DE-AC52-06NA25396) and Sandia National Laboratories (Contract DE-AC04-94AL85000).

■

REFERENCES

(1) Bhat, M. K.; Bhat, S. Cellulose degrading enzymes and their potential industrial applications. Biotechnol. Adv. 1997, 15, 583−620. (2) Somerville, S.; Young, H.; Taylor, C.; Davis, S. C.; Long, S. P. Feedstocks for lignocellulosic biofuels. Science 2010, 13, 790−792. (3) Zhang, Y. H. P.; Lynd, L. R. A functionally based model for hydrolysis of cellulose by fungal cellulase. Biotechnol. Bioeng. 2006, 94, 888−898. (4) Levine, S. E.; Fox, J. M.; Blanch, H. W.; Clark, D. S. A mechanistic model of the enzymatic hydrolysis of cellulose. Biotechnol. Bioeng. 2010, 107, 37−51. (5) Zhang, Y. H. P.; Himmel, M. E.; Mielenz, J. R. Outlook for cellulase improvement: Screening and selection strategies. Biotechnol. Adv. 2006, 24, 452−481. (6) Moran-Mirabal, J. N.; Santhanam, N.; Corgie, S. C.; Craighead, H. G.; Walker, L. P. Immobilization of cellulose fibrils on solid substrates for cellulase-binding studies through quantitative fluorescence microscopy. Biotechnol. Bioeng. 2008, 101, 1129−1141. (7) Eriksson, J.; Malmsten, M.; Tiberg, F.; Callisen, T. H.; Damhus, T.; Johansen, K. S. Model cellulose films exposed to H. insolens glucoside hydrolase family 45 endo-cellulase - the effect of the carbohydrate-binding module. J. Colloid Interface Sci. 2005, 285, 94− 99. (8) Turon, X.; Rojas, O. J.; Deinhammer, R. S. Enzymatic kinetics of cellulose hydrolysis: A QCM-D study. Langmuir 2008, 24, 3880− 3887. (9) Josefsson, P.; Henriksson, G.; Wagberg, L. The physical action of cellulases revealed by a quartz crystal microbalance study using ultrathin cellulose films and pure cellulases. Biomacromolecules 2008, 9, 249−254. (10) Hook, F.; Kasemo, B.; Nylander, T.; Fant, C.; Sott, K.; Elwing, H. Variations in coupled water, viscoelastic properties, and film thckness of a Mefp-1 protein film during adsorption and cross-linking: A quartz crystal microbalance with dissiption monitoring, ellipsometry, and surface plasmon resonance study. Anal. Chem. 2001, 73, 5796− 5804. (11) Cheng, G.; Liu, Z. L.; Murton, J. K.; Jablin, M.; Dubey, M.; Majewski, J.; Halbert, C.; Browning, J.; Ankner, J.; Akgun, B.; Wang, C.; Esker, A. R.; Sale, K. L.; Simmons, B. A.; Kent, M. S. Neutron reflectometry and QCM-D study of the interaction of cellulases with films of amorphous cellulose. Biomacromolecules 2011, 12, 2216−2224. (12) Kontturi, E.; Suchy, M.; Penttila, P.; Jean, B.; Pirkkalainen, K.; Torkkeli, M.; Serimaa, R. Amorphous characteristics of an ultrathin cellulose film. Biomacromolecules 2011, 12, 770−777. (13) Boisset, C.; Borsali, R.; Schulein, M.; Henrissat, B. Dynamic light scattering study of the two-domain structure of Humicola insolens endoglucansse V. FEBS Lett. 1995, 376, 49−52. 8358

dx.doi.org/10.1021/la300955q | Langmuir 2012, 28, 8348−8358