Available online at www.sciencedirect.com

ScienceDirect Advances in solid-state NMR of cellulose Marcus Foston Nuclear magnetic resonance (NMR) spectroscopy is a wellestablished analytical and enabling technology in biofuel research. Over the past few decades, lignocellulosic biomass and its conversion to supplement or displace non-renewable feedstocks has attracted increasing interest. The application of solid-state NMR spectroscopy has long been seen as an important tool in the study of cellulose and lignocellulose structure, biosynthesis, and deconstruction, especially considering the limited number of effective solvent systems and the significance of plant cell wall three-dimensional microstructure and component interaction to conversion yield and rate profiles. This article reviews common and recent applications of solid-state NMR spectroscopy methods that provide insight into the structural and dynamic processes of cellulose that control bulk properties and biofuel conversion. Addresses Washington University in St. Louis, Department of Energy, Environmental & Chemical Engineering, One Brookings Drive, St. Louis, MO 63130, USA Corresponding author: Foston, Marcus ([email protected], [email protected])

Current Opinion in Biotechnology 2014, 27:176–184 This review comes from a themed issue on Energy biotechnology Edited by Arthur J Ragauskas and Korneel Rabaey

Available online XXX 0958-1669/$ – see front matter, # 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.copbio.2014.02.002

Introduction In light of the evitable depletion of non-renewable feedstocks such as petroleum and the negative impact that petroleum production and utilization has on the global environment, lignocellulosic biomass (e.g. agricultural/ forestry waste residues and non-food grassy/woody dedicated energy crops) has been identified as an abundant renewable resource for producing the global supply of energy and material [1,2]. A proposal central to future lignocellulosic biomass utilization is the development of integrated ‘biorefineries’. A biorefinery would facilitate a conversion process in a fashion analogous to modern petroleum refineries, producing foods, fuels, chemicals, feeds, materials, heat, and power. Though significant research has focused on developing biorefineries based on the conversion of lignocellulosic biomass, to-date large-scale and commercial utilization of lignocellulosic Current Opinion in Biotechnology 2014, 27:176–184

biomass to produce fuels and materials has not occurred. This inability to commercialize biomass utilization on a large-scale economically is, at least in part, due to the inherent complexity of the plant cell wall and the related knowledge gaps associated with the mechanisms of plant cell wall biosynthesis and deconstruction. The complexity of the plant cell wall can be largely attributed to the structure of lignocellulosic biomass. Lignocellulosic biomass is a highly complex, heterogeneous, and hierarchical network of secondary plant cell wall biopolymers, primarily, cellulose, hemicellulose, and lignin [3]. Lignin is an irregular, heterogeneous, threedimensional, and cross-linked polypropenyl phenol biopolymer that has been described as covalently linked to polysaccharides through lignin–carbohydrate linkages (LCC). The major monomeric units comprising lignin includes: p-hydroxyphenyl (H), guaiacyl (G), and syringyl (S) units, having varying degrees of methoxylation. Hemicelluloses are polysaccharides comprised of a variety of 5carbon and 6-carbon sugars. These hemicelluloses are described as cell wall biopolymers with relatively short chain lengths and highly branched chain topology. Lastly, cellulose, the most abundant plant cell wall biopolmyer, firstly, is linear; secondly, has a relatively high degree of polymerization; thirdly, can form both crystalline and amorphous morphologies, and fourthly, contains only b-(1!4) linked D-glucopyranosyl (glucose) units. Currently, the conversion of lignocellulose to a wide-array of fuels and chemicals is dependent on biological routes that target cellulose. This lignocellulose-to-fuel bioconversion process typically includes: firstly, biomass sizereduction; secondly, chemical pretreatments to prepare the biomass for biological deconstruction; thirdly, enzymatic hydrolysis to deconstruct cell wall carbohydrates to fermentable sugars; fourthly, fermentation to a final product; and fifthly, final product separation/purification. There is substantial evidence that indicates the: firstly, chemical and molecular structures of the plant cell wall; secondly, ultrastructure of cellulose; thirdly, interactions of cellulose with other plant cell wall components, and fourthly, 3D microstructure of the plant cell wall effect cellulose deconstruction and sugar yield [4–7]. When the aforementioned factors inhibit cellulose deconstruction, they form a composite property of lignocellulose, known as ‘biomass recalcitrance’ [4–7]. Biomass recalcitrance is typically defined as the inherent ability of biomass to resist deconstruction [4]. In order to manipulate this biomass recalcitrance and ultimately lead to a paradigm shift from petroleum-based www.sciencedirect.com

Solid-state NMR of cellulose Foston 177

refineries to biorefineries, research must focus on: firstly, understanding lignocellulose cell wall biosynthesis and structure; secondly, developing more time, energy, and material efficient deconstruction and conversion technologies; and thirdly, continuing improvement of analytical techniques to support or enable these tasks. This review, related to the later research focus, concentrates on the use of solid-state NMR on cellulose as an enabling technology in biofuels research. There are numerous methodologies or enabling technologies adapted and designed for biofuels research to characterize the chemical, molecular, and ultrastructural features of lignocellulosic biomass. Many of these techniques are outlined in a review by Foston et al. [8] and a book chapter by Beecher et al. [9]. Most traditional methods rely on the separation and isolation of the cell wall components followed by wet chemical characterization. However, though well-established and useful, these methods are not only time and energy consuming, but also invariably alter the native structure of cellulose and lack the ability to characterize the cell wall network and key interactions between cellulose, hemicellulose, and lignin. As a result, more recent emphasis in the area of lignocellulosic biomass and cellulose characterization has relied on diffraction (e.g. electron, X-ray, and neutron) [10,11,12,13,14,15] and spectroscopy (e.g. Fourier transform infrared-red, Raman, and NMR) [8,12,13,14] ,13,14] techniques. After over 40 years of research, the exact structure (ultrastructure) of cellulose within the native plant cell wall is still not fully understood; however, extensive reviews by Atalla et al. [13,14], O’Sullivan et al. [15], and more recently, Kim et al. [12], provide a brief description of what is known about the structure of cellulose and more importantly review the basic principles, capabilities, and limitations of the methods used to characterize cellulose structure. In part, the difficulty in analyzing the structure of cellulose is related to the inability to synthesize cellulose in the lab, requiring isolation from biological producers such as plants, heterotrophic bacteria (e.g. Gluconacetobacter xylinus), slime molds (e.g. Dictyostelium discoideum), and tunicates (an animal, e.g. Halocynthia roretzi) [12,16]. Vascular plants utilize a rosette of protein complexes at the plasma membrane to synthesize the most biofuel relevant cellulose in structures known as cellulose microfibrils (20– 25 nm in width). Intra-chain and inter-chain hydrogen bonding of cellulose chains within that microfibril determines cellulose ordering and interaction with other cell wall components. A result of this biosynthesis, inherent structural differences in cellulose occur that depend on the biological source of that cellulose. Moreover, analyzing the structure of cellulose within an intact plant cell wall using diffraction and/or spectroscopy is also problematic due to the presence of other cell wall components. Thus, the cellulose isolation method used and its effect on cellulose further complicates developing an accurate model for the structure of cellulose. www.sciencedirect.com

Solid-state NMR spectroscopy Solid-state NMR spectroscopy methods can provide not only chemical information but also chemical environment and ultrastructural details that are not easily accessible by other non-destructive high-resolution spectral techniques [17,18]. This makes solid-state NMR methodologies particular useful when studying structural problems in complex biological systems such as elucidating the mechanisms of biosynthesis and deconstruction for lignocellulosic biomass or individual plant cell wall components (i.e. cellulose) [19]. There are a variety of interesting applications of modern solid-state NMR spectroscopy to the study of lignin (many of which were developed for investigations of humic acids) [20–26], of hemicellulose [27–29], of intact lignocellulose [30,31,32,33,34,35], and of, the focus of this review, cellulose. Modern solidstate NMR techniques useful to the study of lignocellulosic biomass include but not limited to: 1D 13C crosspolarization magic angle spinning (CP/MAS), 2D 13C correlation, relaxometry, diffusometry, and spin-diffusion techniques. Relaxometry NMR

In a NMR experiment, nuclear excitation is followed by nuclear relaxation. That relaxation behavior (i.e. rate and function) can be used to fingerprint the microscopic dynamics of nuclei being observed, and defines a NMR relaxometry study [17,18]. NMR relaxometry on cellulose and cell wall polysaccharides has been used to investigate moisture content and the effect of hydration [36], examine the connectivity between lignin and carbohydrates [37], examine the structure of and interactions between plant primary cell wall polysaccharides [38], probe C6-carbon conformation/mobility [39], analyze degradation [40], confirm crystallinity, etc. A 2D-WISE (WIde-line SEparation) experiment separates the relaxation behavior of protons through magnetization transfer to adjacent carbons [18]. Hediger et al. used a 2D-WISE variant, a J-WISE experiment which ensures a very selective detection of molecular motion, to discount oversimplified models of the structure of cellulose microfibrils (isolated from onion primary cell walls) that suggest hydrated cellulose is only at the surface of the microfibril [41]. Diffusometry NMR

There is also large collection of solid-state NMR studies which focus on the observation of the NMR signal originating from an absorbed probe molecule (typically water) to yield information on the structure of the lignocellulose and cellulose. Both NMR relaxometry and diffusometry (experiments designed to determine the dynamic behavior of liquids) measurements of that adsorbed water can be used to provide information about the cell wall pore volume to pore surface area ratio [42–47], pore shape [48,49], paramagnetic impurities, moisture content, degradation [50,51], etc. We must note, diffusometry is not Current Opinion in Biotechnology 2014, 27:176–184

178 Energy biotechnology

necessarily a solid-state NMR method; however, the techniques and equipment used are very similar. 1D

13

C CP/MAS NMR

13

C CP/MAS solid-state NMR experiments are actually the combination of three techniques: firstly, cross polarization; secondly, magic angle spinning; and thirdly, high power decoupling [17,18]. During cross polarization, radio frequency pulses are applied on the proton and carbon channels, transferring magnetization from protons to carbon for detection on the carbon channel. This enhances the intensity of the 13C signals, which are typically low due to a low natural abundance and gyromagnetic ratio, by magnetization transfer from nuclei like 1H with a high natural abundance and gyromagnetic ratio. Cross polarization also circumvents issues with inherently long relaxation times for carbon. Magic angle (54.748 with respect to the direction of the magnetic field) spinning is a technique based on orientation averaging that produces solution NMR like spectra by removing chemical shift anisotropy, while high power 1H decoupling is a technique that can substantially reduce line-broadening by isolating 13C from 1H spin systems during detection. Some of the earliest studies using 13C CP/MAS solid-state NMR on cellulose involved Atalla et al. [13] and Vanderhart et al. [52–54] in studies on cellulose isolated from various sources (i.e. Acetobacter, Valonia, Kraft pulp, and low-DP acid-hydrolyzed cellulose) where carbon chemical shifts for the C4 and C6 ring positions in the anhydroglucose unit changed depending upon the cellulosic source. Researchers concluded this observed variation in chemical shifts occurred because solid-state NMR is sensitive to the magnetic non-equivalences in an environment of chemically equivalent nuclei (crystalline and amorphous cellulose have been shown to generate different chemical shifts) and different cellulosic sources have variations in cellulose crystallinity and crystal lattice structure [13].

A typical 13C CP/MAS NMR spectrum of isolated cellulose has board resonances that extend over chemical shift ranges of d  102–108, 80–92 and 57–67 ppm attributed to the C1, C4, and C6 signals (modeled as Gaussian or Lorentzian functions) of cellulose, respectively (see Figure 1a) [13]. Kono et al., using 13C labeled cellulose biosynthesized by Acetobacter xylinum from a culture medium containing D-[1,3-13C]glycerol or D-[2-13C]glucose, report precise chemical shift assignments for cellulose I allomorphs, cellulose Ia and Ib [37]. Cellulose Ia and Ib are the two naturally occuring crystalline cellulose polymorphs. The cellulose Ia (the meta-stable triclinic onechain crystal structure) polymorph is typically assoicated with algae, bateria, and lower plants, while higher plants contain mostly cellulose Ib (the more thermodynamic favored monoclinic two-chain crystal structure). The thermodynamic stablity of cellulose Ib over cellulose Ia is quite evident in the 13C CP/MAS NMR spectra of Current Opinion in Biotechnology 2014, 27:176–184

cellulose following annealing (at 260–2808C) in various polar media [55–59]. In an effort to elucidate the mechanisms of morpholoical transformations of cellulose crystals into non-native forms, 13C CP/MAS NMR spectrum of cellulose II, IIII, IIIII, and IVI (produced upon treatment of crystalline cellulose with an reagent) were acquired and the 13C chemical shifts for the C1, C4, and C6 were denoted [60]. 13C chemical shifts for cellulose derivatives (e.g. cellulose trinitrate, triacetate, tripropionate, and tributyrate; methyl cellulose; ethyl cellulose; hydroxyl ethyl cellulose; carboxymethylcellulose; cyanoethylcellulose; carbamoylethyl cellulose) can be found in work by Hoshino et al. [61]. The C4 peak in the carbon spectrum of isolated cellulose I is the most commonly utilized peak used to extract ultrastructural information (% crystallinity). C4 cellulose amorphous carbons are represented by a fairly broad signal from d  80–85 ppm, while C4 cellulose crystalline carbons generate a sharper resonance from d  85– 92 ppm (see Figure 1a) [13]. Researchers have since applied various data processing methods including basic peak integrations, multiple peak non-linear least-squared fits, and partial least-squares models to determine cellulose crystallinity. These various data processing methods have also been used to estimate cellulose Ia and cellulose Ib contents, deconvolute contributions of para-crystalline cellulose (described as having more order than amorphous cellulose and less order than crystalline cellulose), and evaluate the percentage of cellulose at accessible versus inaccessible cellulose microfibril surfaces (see Figure 1b) [62–70,71,72]. On the basis of NMR relaxometry and molecular dynamic simulations, Bergenstrhle et al. suggest the characteristic doublet attributed to accessible cellulose microfibril surfaces is due to C4-carbons located on top of two different crystallographic planes [73]. Figure 1c depicts a common model used to describe the spatial distribution of cellulose morphologies within a cellulose microfibril, where the cellulose microfibril has a core–shell configuration and the outer amorphous shell is composed of accessible and inaccessible cellulose microfibril surfaces. Using high-resolution atomic force microscopy (AFM), Ding et al. proposed a primary cell wall model for maize that supports these NMR results and includes an inner core of six true crystalline chains [74]. Ding et al. then proposed that this core is surrounded by 12 para-crystalline (or sub-crystalline) chains, followed by another shell composed of 18 more non-crystalline (amorphous or disordered) chains at the cellulose microfibril surface. A series of studies also indicate lateral dimensions of cellulose crystallites and crystallite bundles can be estimated using the relative intensity of C4 amorphous resonances (i.e. peaks attributed to accessible and inaccessible fibril surfaces) [75–77]. Lastly, there are several spectral editing techniques that take advantage of the fact hemicellulose, lignin, and amorphous/crystalline cellulose has differing relaxation behavior [78,79]. www.sciencedirect.com

Solid-state NMR of cellulose Foston 179

Figure 1

(a)

(c)

Amorphous Inaccessible Surface

C2,3,5 4 O HO

C1

5 3

O 2

1

OH

C4

110

100

Crystalline Core

OH

6

90

C6

80

70

Inaccessible fibril surfaces

Para-crystalline

(b)

Cellulose Iβ

Cellulose Iα+β

60 PPM

92

Para-Crystalline

Accessible fibril surface

Cellulose Iα

Amorphous Accessible Surface

90

88

86

84

82

80 PPM Current Opinion in Biotechnology

Using 1D 13C cross-polarization magic angle spinning (CP/MAS) a high-resolution spectrum of isolated cellulose is obtained. A partial least-squares model incorporating overlapping C4 peaks attributed to cellulose Ia, cellulose Ib, cellulose Ia+b, para-crystalline cellulose, and accessible and inaccessible amorphous cellulose is used to extract ultrastructural information. (a) A representative 13C CP/MAS spectrum of solid isolated cellulose. Basic 2-peak integration is used on the C4 region of a 13C CP/MAS spectrum of isolated cellulose to calculate cellulose % crystallinity. (b) The nonlinear, least-squared, spectral fitting of the C4 region of 13C CP/MAS spectrum of isolated cellulose. (c) The proposed structure of cellulose within the cellulose microfibril.

Park et al. proposed a data processing method to determine cellulose crystallinity that utilized the subtraction of the spectrum of a standard amorphous cellulose [80]. In this study and subsequent work by the same team, cellulose crystallinity index measurement techniques, mainly X-ray diffraction and solid-state 13C NMR, were compared and though different measurement techniques and data processing methods provided different absolute values of crystallinity for different cellulosic sources, the different methodologies produced almost identical trends for crystallinity in those same cellulosic sources [80,81]. In an effort to generate a NMR-based technique to estimate cellulose polymorphism and crystallinity that is a rapid and reliable alternative to the curve-fitting methods described above, Rondeau-Mouro et al. applied chemometric analysis of 13C CP/MAS NMR spectra acquired by varying the contact time of the cross-polarization process. Proton spin-diffusion time values were used to estimate a crystallinity index and the lateral dimensions of cellulose crystallites [82]. In addition, principal component analysis (PCA) models seem to accurately classify cellulosic materials by degree of www.sciencedirect.com

crystallinity, whereas, a multi-linear regression method with the constraints of positivity model allowed a crystallinity index to be calculated, as well as estimating the proportion that is cellulose Ib. Similarly, Okushita et al. used variable contact time 13C CP/MAS experiments and PCA to extract cellulose spectral peak component information and evaluate degree of crystallinity in bacterial cellulose [83]. 13

C CP/MAS NMR experiments, and more generally almost all NMR experiments involving 13C nuclei, benefit significantly from efficient and stable 13C isotopic labeling, increasing signal-to-noise and expanding the library of advanced NMR sequences that can be applied. 13 C enriched lignocellulosic biomass can easily and cheaply (compared to synthetic polymers) be generated via incorporation by photosynthesis in a 13C enriched CO2 environment [84–90]. Almost 95-fold enhancements (with respect to natural abundance) in 13C isotope levels have been verified in corn stover (a viable biofuel feedstock) by combustion analysis and mass spectrometry of evolved CO2 [91]. Current Opinion in Biotechnology 2014, 27:176–184

180 Energy biotechnology

Spin-diffusion NMR

A well-applied solid-state NMR technique for the determination of inter-macromolecular and intra-macromolecular distances is spin-diffusion NMR [18,30,31]. For example, Masuda et al. performed 13C and 1H spindiffusion experiments in an effort to investigate the spatial distribution of amorphous and crystalline regions in bacterial cellulose [92]. It was determined that amorphous C4carbons are more than likely located at distances less than about 1 nm from the crystalline C4-carbons and that those amorphous C4-carbons are not solely localized to the microfibril surface. Contradictory NMR spin-diffusion results were published by Fernandes et al. that seem to support a core–shell model for the cellulose microfibril (isolated from spruce wood), where the core is more ordered than the shell [39]. This contradiction may be due to the fact most plant cellulose synthases are suspected to perform a one-step polymerization reaction, whereas bacterial cellulose may be synthesized in multiple steps. It has been proposed the relationship between polymerization and crystallization rates may be important in determining the resulting structure of cellulose within the microfibril [16]. Foston et al. performed an interesting spin-diffusion NMR study that indicated lignin within intact lignocellulose is in close proximity to hemicellulose followed by amorphous and finally crystalline cellulose (see Figure 2). The rates of magnetization transfer suggests that the plant cell wall structure is intimately blended with no major component greater than 2 nm from one another. Moreover, that the cellulose microfibril has a core–shell configuration, with the amorphous shell being 1 nm thick around the crystalline core [30]. Hong et al. completed an interesting set of dynamic nuclear polarization (DNP) enhanced 13C spin diffusion experiments to determine the structure of protein (expansins) binding to non-crystalline polysaccharides in the primary cell wall of Arabidopsis thaliana [93]. Briefly, DNP NMR relies on efficient electron to proton magnetization transfer and can generate a theoretical 660-fold enhancement in NMR intensity [94]. Thus, it is quite clear the future application of DNP solid-state NMR to biomass has great potential as an advanced biorefinery enabling technology. 2D

13

C correlation NMR

Several more recent studies have reported the 13C and/or 1 H chemical shift assignments for cellulose I, II, and IIII as well as a cellulose–ethylenediamine complex in an effort to elucidate the conformation of the b-D-glucose residues in the cellulose crystallites [95–98]. Many of these studies used through-bond 2D INADEQUATE (Incredible Natural Abundance DoublE QUAntum Transfer Experiment) correlations to facilitate assignment. 2D INADEQUATE experiments yield one-bond correlations via spin–spin, scalar, or J-couplings. J-couplings define the interaction between chemically bonded nuclei in the Current Opinion in Biotechnology 2014, 27:176–184

vicinity of the observed nucleus, containing information about bond distance and angles and providing information on the connectivity of molecules [99,100]. Like 2D INADEQUATE NMR for carbon–carbon correlation, there have also been efforts to develop and apply a similar solid-state 2D NMR technique referred to as MAS-JHMQC spectroscopy to investigate carbon-proton correlations [96]. First applied to disordered solids by Lesage et al. to assign unambiguously carbon resonances [100], 2D INADEQUATE NMR was able to clearly show that chemically identical carbons in cellulose Ia and Ib are magnetically inequalevent and that the primary difference between these two forms of cellulose I is in the conformations of anhydroglucose residues and the torsion angle around the b-1,4-glycosidic linkages [101]. In an application of 2D INADEQUATE NMR on regenerated amorphous cellulose, observed spectral correlations along with candidate local structures for amorphous cellulose and their simulated chemical shift details were used to elucidate possible structure in amorphous cellulose [102]. 2D INADEQUATE along with other 2D and 3D correlation NMR experiments were successfully applied to investigate the dynamic environment of the polysaccharides in the primary cell wall [38]. A 2D 13C–13C spin-exchange NMR experiment using a mixing time with a radio frequency-driven recoupling (RFDR) via a rotor-synchronized p-pulse sequence can be applied to uniformly 13C-enriched cellulose in order to obtain interatomic distance information in the cellulose crystal [103]. 2D INADEQUATE NMR experiments of cellulose II show cellulose II unit cells are filled with two kinds of b-D-glucose residues (A and B) [96,103]. On the basis of RFDR spectra, the interatomic distances from each C1-carbon to the other carbon nuclei were determined, ultimately suggesting cellulose II crystals are composes of two independent chains that have A–A and B–B repeating units and no A–B repeating units [103]. Similarly, in another study 2D INADEQUATE NMR experiments indicate cellulose Ia is composed on two kinds of b-D-glucose residues (A and A0 ) and cellulose Ib is composed on two kinds of b-D-glucose residues (B and B0 ). RFDR spectra of cellulose Ia was used to specify interatomic distances from the C4- and C3-carbons of residue A to the C2- and C4-carbons of residue A0 , respectively [104]. The RFDR spectra could not extract interatomic distances for cellulose Ib, but did reveal that cellulose Ia to be composed of only A–A0 repeating units and that cellulose Ib is composed two independent chains that have B–B and B0 –B0 repeating units and no B–B0 repeating units [104].

Conclusion This review has only covered a limited portion of the published studies applied to the solid-state NMR of biomass. For example, several studies focus specifically on the effect of thermochemical treatments such as dilute www.sciencedirect.com

Solid-state NMR of cellulose Foston 181

Figure 2

τ1

at m

at

rix

se –li gn in

m se –li gn in

ice llu lo

π/2

Tm

τ2

spin diffusion spin diffusion ice llu lo

CP

π/2

DD

m

π/2 π/2 13C

DARR

DD

m

CP

he

π/2 1H

~ 5-8 nm

(b)

DARR = dipolar assisted rotational resonance, broadband recoupling

rix

(a)

he

n cellulose microfibril SELDOM – selects carbons at ~150 and 65 ppm

OH

(c)

1.0

(d)

Normal CP

A1

A2

Acetyl methoxy

A3

G

0.8

C2,3,5 C4a C4c C6

I/Imax

C1

Acetyl

O

OH

OCH H CO 3 3

O

OCH 3

S A3

0.6 0.4

A2

0.2

Tm (ms)

A1 0.0 0

4950 495

100

200 300 Tm(ms)

400

500

(e) 1.0 0.8 I/Imax

61 31

0.6 0.4

crystalline C4 amorphous C4 C2,3,5

0.2

.099 PPM 180

160

140

120

100

80

60

40

20

0.0

0

1000

2000

3000

4000

Tm(ms) Current Opinion in Biotechnology

Using a solid-state NMR technique referred to as a 13C CP SELDOM (SELectivity by DestructiOn of Magnetization) with a mixing time delay for spin diffusion, selective excitation of specific aromatic lignin carbons can be accomplished and the average spatial dimensions between major biopolymers within the plant cell wall can be estimated. (a) A schematic representation illustrating the 13C SELDOM spin-diffusion NMR (p/2 = 908 pulse; CP = cross polarization; DD = decoupling; DARR = dipolar assisted rotational resonance, broadband recoupling; Tm = mixing time). (b) A model depicting the magnetization transfer process. (c) Normal CP (top) and 13C selective excitation and spin diffusion experiments of 13C enriched corn stover stem. Various chemical moieties will be denoted by: A1 carbons, which correspond to aromatic lignin carbons appearing 153 ppm or the S3,5 and G3,4 carbons, A2 carbons corresponding to aromatic lignin carbons appearing 135 ppm or the S1,4 and G1 carbons and A3 carbons attributed to S2,6 and G2,5,6 at 110 ppm. Plots of normalized and T1 adjusted peak intensities of the (d) A1 (red), A2 (green) and A (blue) resonance lines shown in Figure 1c against Tm values and (e) cellulose C2,3,5 (red), C4a (blue) and C4c (green). The solid lines in the right plot only serve to follow trends.

acid pretreatments [105–108] or pyrolysis [109–113] on biomass. However, the utility of solid-state NMR to future biomass and biofuel research is quite rich, and is mainly based on the fact solid-state NMR has a unique combination of advantages including being non-destructive, non-invasive, and selective, but most importantly, the ability to provide not only chemical information but also chemical environment and ultrastructural details.

References and recommended reading Papers of particular interest, published within the period of review, have been highlighted as:

2.

Ragauskas AJ, Williams CK, Davison BH et al.: The path forward for biofuels and biomaterials. Science 2006, 311: 484-489.

3.

Fengel F, Wegener G: Wood: Chemistry, Ultrastructure, Reaction. Walter de Gruyter; 1984.

4.

Himmel M, Ding S, Johnson DK, Adney W, Nimlos M et al.: Biomass recalcitrance: engineering plants and enzymes for biofuels production. Science 2007, 315: 804-807.

5.

Yang B, Dai Z, Ding S, Wyman CE: Enzymatic hydrolysis of cellulosic biomass. Biofuels 2011, 2:421-450.

6.

Zhao X, Zhang L, Liu D: Biomass recalcitrance. Part I: the chemical compositions and physical structures affecting the enzymatic hydrolysis of lignocellulose. Biofuels Bioprod Bioref 2012, 6:465-482.

7.

Mansfield S, Mooney C, Saddler J: Substrate and enzyme characteristics that limit cellulose hydrolysis. Biotechnol Prog 1999, 15:804-816.

 of special interest  of outstanding interest 1.

Pu Y, Zhang D, Singh PM, Ragauskas AJ: The new forestry biofuels sector. Biofuels Bioprod Bioref 2008, 2:58-73.

www.sciencedirect.com

Current Opinion in Biotechnology 2014, 27:176–184

182 Energy biotechnology

8. 

Foston M, Ragauskas AJ: Biomass characterization: recent progress in understanding biomass recalcitrance. Ind Biotechnol 2010, 8:191-208. This review covers the recent advances in analytical methodology to characterize the chemical and molecular features related to the ability of biomass to resist biological deconstruction

9.

Beecher J, Hunt C, Zhu JY: Tools for the characterization of biomass at the nanometer scale. In The Nanoscience and Technology of Renewable Biomaterials. Edited by Lucia L, Rojas O. Wiley–Blackwell; 2009:61-90.

10. Xu F, Shi YC, Wang D: X-ray scattering studies of lignocellulosic biomass: a review. Carbohydr Polym 2013, 94:904-917. 11. Langan P, Evans BR, Foston M, Heller WT, O’Neill H et al.: Neutron technologies for bioenergy research. Ind Biotechnol 2012, 8:209-216. 12. Kim SH, Lee CM, Kafle K: Characterization of crystalline  cellulose in biomass: basic principles, applications, and limitations of XRD, NMR, IR, Raman, and SFG. Korean J Chem Eng 2013, 30:2127-2141. This review covers the analytical methodology to characterize the structure of cellulose. 13. Atalla RH: Celluloses. In Comprehensive Natural Products Chemistry. Edited by Barton D, Nakanishi K, Meth-Cohn O. Elsevier; 1999:529-595. 14. Atalla RH, Isogai A: Recent developments in spectroscopic and chemical characterization of cellulose. In Polysaccharides: Structural Diversity and Functional Versatility. Edited by Dumitriu S. CRC Press; 2010:123-157. 15. O’Sullivan AC: Cellulose: the structure slowly unravels.  Cellulose 1997, 4:173-207. This review described work detailing the structure of cellulose. 16. Saxena IM, Brown M Jr: Cellulose biosynthesis: current views and evolving concepts. Ann Bot 2005, 9:9-21. 17. Duer MJ: Introduction to Solid-State NMR Spectroscopy. Oxford: Blackwell; 2004, . 18. Schmidt-Rohr K, Spiess HW: Multidimensional Solid-state NMR and Polymers. Academic Press; 1994. 19. Asakura T, Ando I: Solid-State NMR of Polymers. Elsevier Science; 1998. 20. Hatcher P: Chemical strcutural studies of natural lignin by dipolar dephasing solid-state 13C nuclear magnetic resonance. Org Geochem 1987, 11:31-39. 21. Holtman KM, Chen N, Chappell MA, Kadla JF: Chemical structure and heterogeneity differences of two lignins from loblolly pine as investigated by advanced solid-state NMR spectroscopy. J Agric Food Chem 2010, 58:9882-9892. 22. Mao JD, Xing B, Schmidt-Rohr K: New structural information on a humic acid from two-dimensional 1H–13C correlation solidstate nuclear magnetic resonance. Environ Sci Technol 2001, 35:1928-1934. 23. Mao JD, Holtman KM, Scott JT, Kadla JF, Schmidt-Rohr K: Differences between lignin in unprocessed wood, milled wood, mutant wood, and extracted lignin detected by 13C solid-state NMR. J Agric Food Chem 2006, 54:9677-9686. 24. Schmidt-Rohr K, Mao JD: Efficient CH-group selection and identification in 13C solid-state NMR by dipolar DEPT and 1H chemical-shift filtering. JACS 2002, 124:13938-13948. 25. Liitia¨ T, Maunu SL, Sipila J, Hortling B: Application of solid-state 13 C NMR spectroscopy and dipolar dephasing technique to determine the extent of condensation in technical lignins. Solid State Nucl Magn Reson 2002, 21:171-186. 26. Manders WF: Solid-state 13C NMR determination of the syringyl/ guaiacyl ratio in hardwoods. Holzforschung 1987, 41:13-18. 27. Komatsu T, Kikuchi J: Selective signal detection in solid-state NMR using rotor-synchronized dipolar dephasing for the analysis of hemicellulose in lignocellulosic biomass. J Phys Chem Lett 2013, 4:2279-2283. Current Opinion in Biotechnology 2014, 27:176–184

28. Liitia¨ T, Maunu SL, Hortling B, Tamminen T et al.: Cellulose crystallinity and ordering of hemicelluloses in pine and birch pulps as revealed by solid-state NMR spectroscopic methods. Cellulose 2003, 10:307-316. 29. Teleman A, Larsson PT, Iversen T: On the accessibility and structure of xylan in birch kraft pulp. Cellulose 2001, 8:209-215. 30. Foston M, Katahira R, Gjersing E, Davis M, Ragauskas AJ: Solidstate selective 13C excitation and spin diffusion NMR to resolve spatial dimensions in plant cell walls. J Agric Food Chem 2012, 60:1419-1427. 31. Bardet M, Emsley L, Vincendon M: Two-dimensional spinexchange solid-state NMR studies of 13C-enriched wood. Solid State Nucl Magn Reson 1997, 8:25-32. 32. Liitia¨ T: Application of Modern NMR Spectroscopic Techniques to  Structural Studies of Wood and Pulp Components. Department of Chemistry, University of Helsinki; 2002. This thesis uses high-resolution solid-state NMR spectroscopy in morphological studies of cellulose and in structural studies of lignin and lignincarbohydrate complexes. 33. Gil AM, Neto CP: Solid-state NMR studies of wood and other  lignocellulosic materials. Annu Rep NMR Spectrosc 1999, 37:75-117. This review covers the significant ulitization of solid-state NMR applied to the structural characterization of the main components of lignocellulose. 34. Bardet M, Foray MF, Tran QK: High-resolution solid-state CPMAS NMR study of archaeological woods. Anal Chem 2002, 74:4386-4390. 35. Gerasimowicz WV, Hicks KB, Pfeffer PE: Evidence for the existence of associated lignin–carbohydrate polymers as revealed by carbon-13 CPMAS solid-state NMR spectroscopy. Macromolecules 1984, 17:2597-2603. 36. Felby C, Thygesen LG, Kristensen JB et al.: Cellulose–water interactions during enzymatic hydrolysis as studied by time domain NMR. Cellulose 2008, 15:703-710. 37. Ahvazi BC, Argyropoulos DS: Proton spin–lattice relaxation time measurements of solid wood and its constituents as a function of pH: Part II. Sol State Nucl Magn Resonan 1999, 15:4957. 38. Dick-Perez M, Zhang Y, Hayes J, Salazar A, Zabotina OA, Hong M: Structure and interactions of plant cell-wall polysaccharides by two- and three-dimensional magic-angle-spinning solidstate NMR. Biochem 2011, 50:989-1000. 39. Fernandes AN, Thomas LH, Altaner CM, Callow P et al.: Nanostructure of cellulose microfibrils in spruce wood. PNAS 2011, 108:E1195-E1203. 40. Soares S, Ricardo N, Jones S, Heatley F: High temperature thermal degradation of cellulose in air studied using FTIR and 1H and 13C solid-state NMR. Eur Polym J 2001, 37: 737-745. 41. Hediger S, Lesage A, Emsley L: A new NMR method for the study of local mobility in solids and application to hydration of biopolymers in plant cell walls. Macromolecules 2002, 35:50785084. 42. Haggkvist M, Li TQ, Odberg L: Effects of drying and pressing on the pore structure in the cellulose fiber wall studied by 1H and 2 H NMR relaxation. Cellulose 1998, 5:33-49. 43. Haggkvist M: The influence of two wet strength agents on pore size and swelling of pulp fibres and on tensile strength properties. Nord Pulp Pap Res J 1998, 13:292-298. 44. Topgaard D, So¨derman O: Diffusion of water absorbed in cellulose fibers studied with 1H-NMR. Langmuir 2001, 17:26942702. 45. Topgaard D, So¨derman O: Changes of cellulose fiber wall structure during drying investigated using NMR self-diffusion and relaxation experiments. Cellulose 2002, 9:139-147. 46. Li TQ, Henriksson U: Determination of pore sizes in wood cellulose fibers by 2H and 1H NMR. Nord Pulp Pap Res J 1993, 3:326-330. www.sciencedirect.com

Solid-state NMR of cellulose Foston 183

47. Li TQ, Henriksson U, Eriksson JC, Odberg L: Water–cellulose interaction in wood pulp fiber suspensions studied by 170 and 2H NMR relaxation. The effect of beating. Langmuir 1992, 8:680-686. 48. Topgaard D, So¨derman O: Experimental determination of pore shape and size using q-space NMR microscopy in the long diffusion-time limit. Magn Reson Imag 2003, 21:69-76. 49. Topgaard D, So¨derman O: Self-diffusion in two- and threedimensional powders of anisotropic domains: an NMR study of the diffusion of water in cellulose and starch. J Phys Chem B 2002, 106:11887-11892. 50. Blumich B, Anferova S, Sharma S, Federici C: Degradation of historical paper: nondestructive analysis by the NMR-MOUSE. J Magn Reson 2003, 161:204-209. 51. Elder T, Labbe N, Rials T: Time domain-nuclear magnetic resonance study of chars from southern hardwood. Biomass Bioenerg 2006, 30:855-862. 52. VanderHart DL, Atalla RH: Observations by high-resolution carbon-13 nuclear magnetic resonance of cellulose I related to morphology and crystal structure. Macromolecules 1981, 14:570-574. 53. VanderHart DL, Atalla RH: Studies of microstructure in native celluloses using solid-state carbon-13 NMR. Macromolecules 1984, 17:1465-1472. 54. VanderHart DL, Atalla RH: Native cellulose: a composite of two distinct crystalline forms. Science 1984, 223:283-285. 55. Debzi EM, Chanzy H, Sugiyama J, Tekely P, Excoffier G: The Ia ! Ib transformation of highly crystalline cellulose by annealing in various mediums. Macromolecules 1991, 24:68166822. 56. Yamamoto H, Horii F: CPMAS carbon-13 NMR analysis of the crystal transformation induced for valonia cellulose by annealing at high temperatures. Macromolecules 1993, 26:1313-1317. 57. Yamamoto H, Horii F, Odani H: Structural changes of native cellulose crystals induced by annealing in aqueous alkaline and acidic solutions at high temperatures. Macromolecules 1989, 22:4130-4132. 58. Sugiyama J, Okano T, Yamamoto H, Horii F: Transformation of valonia cellulose crystals by an alkaline hydrothermal treatment. Macromolecules 1990, 23:3196-3198. 59. Horii F, Yamamoto H, Kitamaru R, Tanahashi M, Higuchi T: Transformation of native cellulose crystals induced by saturated steam at high temperatures. Macromolecules 1987, 20:2946-2949. 60. Isogai A, Usuda M, Kato T, Uryu T et al.: Solid-state CP/MAS NMR study of cellulose polymorphs. Macromolecules 1989:3168-3172.

13

C

61. Hoshino M, Takai M, Fukuda K, Imura K et al.: 13C-NMR study of cellulose derivatives in the solid state. J Polym Sci A Polym Chem 1989, 27:2083-2092. 62. Mosier SL, Littia T, Kauliomaki S, Hortling B, Sundquist J: 13C CPMAS NMR investigations of cellulose polymorphs in different pulps. Cellulose 2000, 7:147-159. 63. Larsson PT, Westermark U, Iversen T: Determination of the cellulose Ia allomorph content in a tunicate cellulose by CP/ MAS 13C NMR spectroscopy. Carbohydr Res 1995, 278:339-343. 64. Larsson PT, Wickholm K, Iversen T: A CP/MAS 13C NMR investigation of molecular ordering in celluloses. Carbohydr Res 1997, 302:19-25. 65. Larsson PT, Hult EL, Wickholm K, Pettersson E, Iversen T: CP/ MAS 13C-NMR spectroscopy applied to structure and interaction studies on cellulose I. Sol State Nucl Magn Resonan 1999, 15:31-40. 66. Lennholm H, Larsson T, Iversen T: Determination of cellulose Ia and Ib in lignocellulosic materials. Carbohydr Res 1994, 261:119-131. www.sciencedirect.com

67. Hult EL, Larsson PT, Iversen T: A comparative CP/MAS 13C-NMR study of the supermolecular structure of polysaccharides in sulphite and kraft pulps. Holzforschung 2002, 56:179-184. 68. Hult EL, Liitia T, Maunu SL, Hortling B, Iversen T: A CP/MAS 13CNMR study of cellulose structure on the surface of refined kraft pulp fibers. Carbohydr Polym 2002, 49:231-234. 69. Hult EL, Iversen T, Sugiyama J: Characterization of the supermolecular structure of cellulose in wood pulp fibres. Cellulose 2003, 10:103-110. 70. Hult EL, Larsson PT, Iversen T: A comparative CP/MAS 13C-NMR study of cellulose structure in spruce wood and kraft pulp. Cellulose 2000, 7:35-55. 71. Wickholm K, Larsson PT, Iversen T: Assignment of non crystalline forms in cellulose I by CP/MAS 13C NMR spectroscopy. Carbohydr Res 1998, 312:123-129. This study makes new 13C assisngments for cellulose at accessible fibril surfaces, cellulose at inaccessible fibril surfaces, para-crystalline cellulose, and hemicellulose. In addition, the commonly utilized 7-peak nonlinear least-squared fit method was first proposed, including an appilcation to determine cellulose microfibril dimesions. 72. Wickholm K, Hult EL, Larsson PT, Iversen T, Lennholm H: Quantification of cellulose forms in complex cellulose materials: a chemometric model. Cellulose 2001, 8:139-148. 73. Bergenstra˚hle M, Wohlert J, Larsson PT, Mazeau K, Berglund LA: Dynamics of cellulose–water interfaces: NMR spin–lattice relaxation times calculated from atomistic computer simulations. J Phys Chem B 2008, 112:2590-2595. 74. Ding SY, Himmel ME: The maize primary cell wall microfibril: a new model derived from direct visualization. J Agric Food Chem 2006, 54:597-606. 75. Newman RH: Estimation of the lateral dimensions of cellulose crystallites using 13C NMR signal strengths. Sol State Nucl Magn Resonan 1999, 15:21-29. 76. Huex L, Dinand E, Vignon MR: Structural aspects in ultrathin cellulose microfibrils followed by 13C CP-MAS NMR. Carbohydr Polym 1999, 40:115-124. 77. Hult EL, Larsson PT, Iversen T: Cellulose fibril aggregation—an inherent property of kraft pulps. Polymer 2001:42. 78. Newman RH, Hemmingson JA: Determination of the degree of cellulose crystallinity in wood by carbon-13 nuclear magnetic resonace spectroscopy. Holzforschung 1990, 44:351-355. 79. Newman RH, Hemmingston JA: Carbon-13 NMR distinction between categories of molecular order and disorder in cellulose. Cellulose 1994, 2:95-110. 80. Park S, Johnson D, Ishizawa C, Parilla P, Davis M: Measuring the crystallinity index of cellulose by solid state 13C nuclear magnetic resonance. Cellulose 2009, 16:641-647. 81. Park S, Baker JO, Himmel ME, Parilla PA, Johnson DK: Cellulose  crystallinity index: measurement techniques and their impact on interpreting cellulase performance. Biotechnol Biofuels 2010, 3:10. Cellulose crystallinty can be measured via different methods. This study finds the absoulte value of cellulose crystallinty is highly dependent on the measurement technique and data interpretation used, but also, that these differnces in crystallinty generate trends within sample sets that are consistent dispite the measurement technique. 82. Rondeau-Mouro C, Bizot H, Bertrand BD: Chemometric analyses of the 1H–13C cross-polarization build-up of celluloses NMR spectra: a novel approach for characterizing the cellulose crystallites. Carbohydr Polym 2011, 84:539-549. 83. Okushita K, Komatsu T, Chikayama E, Kikuchi J: Statistical approach for solid-state NMR spectra of cellulose derived from a series of variable parameters. Polym J 2012, 44:895-900. 84. Mochida K, Furuta T, Ebana K, Shinozaki K, Kikuchi J: Correlation exploration of metabolic and genomic diversity in rice. BMC Genom 2009, 10:568-578. 85. Sekiyama Y, Kikuchi J: Towards dynamic metabolic network measurements by multi-dimensional NMR-based fluxomics. Phytochemistry 2007, 68:2320-2329. Current Opinion in Biotechnology 2014, 27:176–184

184 Energy biotechnology

86. Sekiyama Y, Chikayama E, Kikuchi J: Profiling polar and semipolar plant metabolites throughout extraction processes using a combined solution-state and high-resolution magic angle spinning NMR approach. Anal Chem 2010, 82:1643-1652. 87. Sekiyama Y, Chikayama E, Kikuchi J: Evaluation of a semipolar solvent system as a step toward heteronuclear multidimensional NMR-based metabolomics for 13C-labeled bacteria, plants, and animals. Anal Chem 2011, 83:719-726. 88. Terashima N, Atalla RH, Vanderhart DL: Solid state NMR spectroscopy of specifically 13C-enriched lignin in wheat straw from coniferin. Phytochemistry 1997, 46:863-870. 89. Tian C, Chikayama E, Tsuboi Y, Kuromori T, Shinozaki K, Kikuchi J, Hirayama T: Top-down phenomics of Arabidopsis thaliana. J Biol Chem 2007, 282:18532-18541.

100. Lesage A, Bardet M, Emsley L: Through-bond carbon–carbon connectivities in disordered solids by NMR. JACS 1999, 121:10987-10993. 101. Kono H, Erata T, Takai M: Determination of the through-bond carbon–carbon and carbon–proton connectivities of the native celluloses in the solid state. Macromolecules 2003, 36:5131-5138. 102. Mori T, Chikayama E, Tsuboi Y, Ishida Y, Shisa N et al.: Exploring the conformational space of amorphous cellulose using NMR chemical shifts. Carbohydr Polym 2012, 90:1197-1203. 103. Kono H, Numata Y: Two-dimensional spin-exchange solidstate NMR study of the crystal structure of cellulose II. Polymer 2004:4541-4547.

90. Kikuchi J, Shinozaki K, Hirayama T: Stable isotope labeling of Arabidopsis thaliana for an NMR-based metabolomics approach. Plant Cell Physiol 2004, 45:1099-1104.

104. Kono H, Numata Y: Structural investigation of cellulose Ia and Ib by 2D RFDR NMR spectroscopy: determination of sequence of magnetically inequivalent D-glucose units along cellulose chain. Cellulose 2006, 13:317-326.

91. Foston M, Samuel R, Ragauskas AJ: 13C cell wall enrichment and ionic liquid NMR analysis: progress towards a highthroughput detailed chemical analysis of the whole plant cell wall. Analyst 2012, 137:3904-3909.

105. Foston M, Ragauskas AJ: Changes in lignocellulosic supramolecular and ultrastructure during dilute acid pretreatment of populus and switchgrass. Biomass Bioenerg 2010, 34:1885-1898.

92. Masuda K, Adachi M, Hirai A, Yamamoto H et al.: Solid-state 13C and 1H spin diffusion NMR analyses of the microfibril structure for bacterial cellulose. Solid State Nucl Magn Reson 2003, 23:198-212. 93. Wang T, Park YB, Caporini MA, Rosay M, Zhong L, Cosgrove DJ, Hong M: Sensitivity-enhanced solid-state NMR detection of expansin’s target in plant cell walls. PNAS 2013, 110:1644416449. 94. Barnes AB, De Paepe G, Van der Wel PCA, Hu KN, Joo CG, Bajaj VS, Griffin RG et al.: High-field dynamic nuclear polarization for solid and solution biological NMR. Appl Magn Reson 2008, 34:237-263. 95. Kono H, Erata T, Takai M: Complete assignment of the CP/MAS 13 C NMR spectrum of cellulose IIII. Macromolecules 2003, 36:3589-3592. 96. Kono H, Numata Y, Erata T, Takai M: 13C and 1H resonance assignment of mercerized cellulose II by two-dimensional MAS NMR spectroscopy. Macromolecules 2004, 37:5310-5316. 97. Numata Y, Kono H, Kawano S, Erata T, Takai M: Crosspolarization/magic-angle spinning 13C nuclear magnetic resonance study of cellulose I–ethylenediamine complex. J Biosci BioEng 2003, 96:461-466. 98. Kono H, Yunoki S, Shikano T, Fujiwara M et al.: CP/MAS 13C NMR study of cellulose and cellulose derivatives. 1. Complete assignment of the CP/MAS 13C NMR spectrum of the native cellulose. JACS 2002, 124:7506-7751. 99. Cedars S, Lesage A, Emsley L: Chemical shift correlations in disordered solids. JACS 2005, 127:4466-4476.

Current Opinion in Biotechnology 2014, 27:176–184

106. Foston M, Ragauskas AJ: Changes in the structure of the cellulose fiber wall during dilute acid pretreatment in populus studied by 1H and 2H NMR. Energy Fuels 2010, 24:5677-5685. 107. Pu Y, Ziemer C, Ragauskas AJ: CP/MAS 13C NMR analysis of cellulase treated bleached softwood kraft pulp. Carbohydr Res 2006, 341:591-597. 108. Samuel R, Pu Y, Foston M: Solid-state NMR characterization of switchgrass cellulose after dilute acid pretreatment. Biofuels 2010, 1:85-90. 109. Falco C, Perez Caballero F, Babonneau F, Gervais C et al.: Hydrothermal carbon from biomass: structural differences between hydrothermal and pyrolyzed carbons via 13C solid state NMR. Langmuir 2011, 27:14460-14471. 110. Cao X, Pignatello JJ, Li Y, Lattao C et al.: Characterization of wood chars produced at different temperatures using advanced solid-state 13C NMR spectroscopic techniques. Energy Fuels 2012, 26:5983-5991. 111. Melkior T, Jacob S, Gerbaud G, Hediger S et al.: NMR analysis of the transformation of wood constituents by torrefaction. Fuel 2012, 92:271-280. 112. Brewer CE, Schmidt-Rohr K, Satrio JA, Brown RC: Characterization of biochar from fast pyrolysis and gasification systems. Environ Progr Sustain Energy 2009, 28:386-396. 113. Wooten JB, Seeman JI, Hajaligol MR: Observation and characterization of cellulose pyrolysis intermediates by 13C CPMAS NMR. A new mechanistic model. Energy Fuels 2004, 18.

www.sciencedirect.com