Many copolysaccharides are compositionally heterogeneous, containing variations in monosaccharide composition and sequence distributions , , , . This heterogeneity may originate from different sources: 1) It can be caused by variations in biosynthesis due to different stages of plant maturation, seasonal variations, or differences in temperature, drought, and nutrient profile during plant growth. 2) Different plant varieties or cultivars may produce copolymers with slightly different composition. 3) The process of extraction of the polysaccharides from a plant or downstream processing may vary, leading to minor differences in composition. 4) If the polysaccharide has been subjected to post-harvest modification (e.g. alkaline hydrolysis, enzymatic reaction, or chemical derivatization), the variations in reaction conditions or process fluctuations may cause changes in composition and sequence distribution. 5) Sometimes a commercial material may be a blend of different batches of materials. In view of the possibility of heterogeneity, NMR analysis of these polysaccharides should be done with care. Frequently the NMR sequence information (e.g., diad and triad sequence intensities) is fitted to simple Bernoullian (B), first-order Markovian (M1) or second-order Markovian (M2) models. For compositionally heterogeneous polymers, the use of these conventional models can be misleading.
A number of NMR methodologies have been developed to address compositional heterogeneity , , , , , , , , , , , , . For simplicity, these treatments can be grouped into two general types: continuous function and discrete-component approaches.
In the continuous function treatment, one can use the perturbed Markovian models, such as symmetric  and non-symmetric functions  or the general (function-independent) formalism . For example, the exponentially modified Gaussian (EMG) function  can be utilized, where the Bernoullian probability is represented not by one value but by a distribution of values, f(z):
where z is the Bernoullian probability, N the area under the Gaussian, σ the standard deviation, τ the skew factor, z′ the dummy variable of integration, and P1 the average value of Bernoullian probability in the absence of exponential modification. The EMG model for the first-order Markovian probabilities can be similarly expressed . For these models, the theoretical expressions for polymer composition, diad, triad, and tetrad sequences have been previously derived . The observed NMR sequence intensities can be fitted to the theoretical expressions to obtain P1, σ, and τ.
In the discrete-component approaches , , , the polymer is regarded as the mixture of two or more discrete components. The components may be separate chains or joined together as a block copolymer. In this approach, the composition and the sequence intensities are the weighted averages of the corresponding compositions and sequence intensities of all the components.
where Iβ is the total intensity for sequence β, wκ is the weight fraction for component κ, and Iβκ is the intensity for sequence β and component κ. In simple cases, the observed and the calculated NMR sequence intensities can be fitted to the two-component B/B or M1/M1 model; if NMR sequence data on polymer fractions are available, it may be possible to combine the NMR data of the fractions and fit them to three- or four-component models.
The methodologies described above were initially developed to treat compositional heterogeneity observed in synthetic polymers , , , , , , , , , . The same approaches have also been shown to be applicable to copolysaccharides , , , , . In this work a review is made of the NMR analysis of three polysaccharides that exhibit compositional heterogeneity.
Chitin is a homopolymer of 2-acetamido-2-deoxy-β-D-glucopyranose (GlcNAc). Chitosan is partially deacetylated chitin and may be considered a copolymer of GlcNAc residue and 2-amino-2-deoxy-β-D-glucopyranose (GlcN) residue .
The 1H and 13C NMR spectra of chitosan have been previously assigned by Varum et al. . They have also reported triad sequence intensities from both heterogeneous and homogeneous deacetylation reactions , , . Their data can be fitted to both conventional (unperturbed) and perturbed Markovian models . The NMR triad data from two chitosan samples, one made via homogeneous reaction, and the other one via heterogeneous reaction, together with analysis, are given in Table 1, where A=acetylated unit (GlcNAc residue) and D=deacetylated unit (GlcN residue).
With the unperturbed M1 model, the mean deviations are relatively large (1.4–3.0%). When the perturbed M1-EMG model is used, the mean deviations decrease substantially, showing a much better fit. Thus, the NMR analysis of these chitosan samples indicates that both are heterogeneous in composition. This finding is consistent with the results of fractionation and other analyses , which also showed chitosan to be compositionally heterogeneous.
Alginates are anionic copolysaccharides isolated from brown algae or produced through bacterial fermentation . They are copolymers of β-D-mannuronate (M) and α-L-guluronate (G) residues. The monomer residues may be arranged in homopolymer, alternating, or random sequences , e.g.
NMR studies on alginates have been well documented in the literature , , . The assignments for both 1H and 13C NMR spectra are known; diad and triad sequence distributions can be readily derived from the NMR spectra , . It was shown earlier that previously published NMR data on alginate ,  could be fitted to perturbed Markovian and/or discrete component models . Furthermore, three commercial samples of sodium alginates were fractionated by size exclusion chromatography (SEC), and several SEC fractions were analyzed by NMR using the same methodology as delineated here . The coupled SEC-NMR data showed that at least two separate components (and perhaps even three) could be identified. The possible linkage of these two components to the enzymatic action in the M→G epimerization reaction and their molecular weight dependence were noted in the earlier report .
For example, the NMR data for two (unfractionated) alginate samples  are given in Table 2, where large mean deviations are observed for the conventional M1 model (around 4%) but mean deviations decrease to <1% with the two-component M1/M1 model.
Commercially available pectin is obtained by extraction from citrus peel, apple pomace, and sugar beet. Pectin has a complex structure, of which about 60% consists of homogalacturonan, with a backbone of α-1,4-linked GalA, which is partly methylated at C-6 . One of the major determinants in pectin properties is the degree of esterification (DE). Sucrose is usually used to form gels with high methoxy pectin (DE>50%), whereas Ca++ is often added for the gelation of low methoxy pectin (DE<50%). Other than being a gelling agent, pectin can also serve as a thickener and stabilizer in food applications .
A typical 13C NMR spectrum of a citrus pectin with a DE of 70% is given in Fig. 1. The assignments , , ,  have been labeled on the spectrum, where G and E refer to galacturonic acid and ester, respectively. The ester carbonyl appears on the spectrum as four peaks corresponding to the E-centered triad sequences (Fig. 1 inset). In addition, C-1, C-4, and C-5 (at 99.6, 78.6 and 71 ppm, respectively) all show splitting due to their sensitivity towards the ester formation and can be used to determine the DE of the sample. Other features of the spectrum include the methyl ester peak at 54.3 ppm, and numerous small peaks from neutral sugars, primarily rhamnose, galactose, arabinose, xylose, and fucose. In addition, the 1H NMR spectrum (not shown) is also rich in sequence information and can be combined with 13C to provide more detailed G/E sequence distributions.
In previous work , , the present author was involved with a research program to use NMR to determine the sequence distribution of G and E residues in a series of pectin samples. For example, the NMR triad data for two fractions of a pectin sample extracted from citrus peel are given in Table 3. These fractions were obtained through a proprietary process, where fraction C2 was expected to contain a more blocky distribution of unesterified GalA units than fraction C1.
The analysis can be carried out with conventional, perturbed, and multicomponent Bernoullian models. The fit with the B model gives relatively large mean deviations (Table 3). The results obtained for the perturbed B-EMG model give smaller mean deviations; the non-zero values obtained for σ and τ indicate that both fractions are heterogeneous in nature. The σ and τ values also permit the chemical composition distribution (CCD) curves to be calculated as shown in Fig. 2.
For the multicomponent model, since both fractions C1 and C2 originate from the same pectin sample, their NMR data can be combined and analyzed simultaneously, thereby increasing the degree of freedom for the analysis. As shown in Table 3, the data can be fitted to a three-component B/B/B model with DE values of 77%, 67% and 2%. Components 1 and 2 are random in methyl ester sequence distribution while component 3 is an almost homopolymer block of galacturonan. As expected, the two fractions C1 and C2 comprise different proportions of these three components, as indicated in Table 3 and (as bar graphs) in Fig. 2. Fraction C2 contains a higher amount of component 3 than fraction C1, indicating a somewhat blockier microstructure.
Thus, in agreement with the current understanding of pectin structure , the NMR data analysis confirms that pectin is compositionally heterogeneous and that it contains blocks of GalA. The combination of NMR and fractionation is very helpful because it permits the three-component model to be tested in this case.
A few comments may be made as to the utility of the NMR methodology. First, at least some of the NMR peaks for a copolysaccharide spectrum need to show splittings due to sequence distribution; this is a prerequisite for NMR sequence analysis. If the splittings are small, higher magnetic field, better spectral resolution, and sometimes higher probe temperatures can improve peak separation. Secondly, the peak intensities should optimally be split to the triad (or higher) sequence level. Thirdly, the precision of the peak intensities must be optimized. It is recommended that peak integration be repeated at least two to three times and an average value taken. Curve deconvolution can also help to obtain more accurate intensities when the peaks are overlapped. Finally, other analytical techniques (such as fractionation, SEC, HPLC, IR, and mass spectrometry) may be combined with NMR to maximize the overall information content. For example, these techniques may offer additional proofs of the heterogeneity of the polysaccharide in question. Alternatively, if the polysaccharide can be separated into different fractions (e.g. through fractionation or chromatography), the NMR data from the fractions can be analyzed simultaneously to provide further information.
This article aims to review the theoretical basis and the methodologies that have been developed for the NMR analysis of heterogeneous polymer systems. Three examples have been given to show how these techniques can be applied to the analysis of NMR data of polysaccharides. In these cases, the sequence intensities have been tested by either the perturbed or multicomponent models for the presence of compositional heterogeneity. As advances in technology enable the magnetic field frequency to increase (now at 1 GHz for 1H NMR), the NMR spectra of more copolysaccharides may show sensitivity to sequence effects, and the sequence intensities may also be measured at greater precision. These developments will hopefully permit more copolysaccharides to be analyzed in the future by the NMR methodologies described herein.
Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture. USDA is an equal opportunity provider and employer.
K. I. Draget, S. T. Moe, G. Skjak-Braek, O. Smidsrod. Alginates, in Food Polysaccharides and their Applications, A. M. Stephen, G. O. Phillips, P. A. Williams (Eds.), pp. 289–334, Taylor and Francis, Boca Raton (2006). Google Scholar
J. F. Ross. J. Macromol. Sci.-Chem. A23, 1451 (1986). Google Scholar
H. N. Cheng. Makromol. Chem.: Theory Simul. 1, 415 (1992). Google Scholar
H. N. Cheng. Makromol. Chem.: Theory Simul. 2, 901 (1993). Google Scholar
H. N. Cheng. Makromol. Chem.: Theory Simul. 2, 561 (1993). Google Scholar
T. G. Neiss, H. N. Cheng, P. J. H. Daas, H. A. Schols. ACS Polym. Prepr. 39(1), 688 (1998). Google Scholar
T. G. Neiss, H. N. Cheng, P. Daas, H. A. Schols. Macromol. Symp. 140, 165 (1991). Google Scholar
H. N. Cheng. ACS Polym. Prepr. 44(1), 261 (2003). Google Scholar
E. Westerlund, P. Aman, R. E. Andersson, R. Andersson. Carbohydr. Polym. 14, 179 (1991). Google Scholar
About the article
Published Online: 2017-01-11
Published in Print: 2017-07-26
Citation Information: Pure and Applied Chemistry, Volume 89, Issue 7, Pages 877–883, ISSN (Online) 1365-3075, ISSN (Print) 0033-4545, DOI: https://doi.org/10.1515/pac-2016-1020.http://creativecommons.org/licenses/by-nc-nd/4.0/.