Subscriber access provided by READING UNIV
Article
Chitosan analysis by enzymatic / mass spectrometric fingerprinting and in silico predictive modeling Anna Niehues, Jasper Wattjes, Julie Bénéteau, Gustavo R. Rivera-Rodriguez, and Bruno M. Moerschbacher Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.7b04002 • Publication Date (Web): 31 Oct 2017 Downloaded from http://pubs.acs.org on November 3, 2017
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 8
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
Chitosan analysis by enzymatic / mass spectrometric fingerprinting and in silico predictive modeling Anna Niehues ‡, Jasper Wattjes ‡, Julie Bénéteau, Gustavo R. Rivera-Rodriguez†, Bruno M. Moerschbacher* Institute for Biology and Biotechnology of Plants, University of Muenster, Schlossplatz 8, 48143 Muenster, Germany ABSTRACT: Chitosans, β-1,4-linked partially N-acetylated linear polyglucosamines, are very versatile and promising functional biopolymers. Understanding their structure-function relationships requires sensitive, accurate structural analyses to determine parameters like degree of polymerization (DP), fraction of acetylation (FA) or pattern of acetylation (PA). NMR, the gold standard for FA analysis, requires large amounts of sample. Here, we describe an enzymatic / mass spectrometric fingerprinting method to analyze the FA of chitosan polymers. The method combines the use of chitinosanase, a sequence-specific hydrolase that cleaves chitosan polymers into oligomeric fingerprints, UHPLC-ESI-MS, and partial least squares regression (PLSR). We also developed a technique to simulate enzymatic fingerprints in silico that were used to build the PLS models for FA determination. Overall, we found our method to be as accurate as NMR while at the same time requiring only microgram amounts of sample. Thus the method represents a powerful technique for chitosan analysis.
Chitin, a linear β-(1,4)-linked polymer composed of N-acetyl-ᴅ-glucosamine (GlcNAc, A) residues, is one of the most abundant natural polysaccharides. It occurs as a structural element in the exoskeletons of crustaceans and insects, in the egg shells of nematodes, in the endoskeletons of mollusks, and in the cell walls of fungi. In fungi, morphogenesis may involve partial or complete de-N-acetylation, converting GlcNAc units to ᴅ-glucosamine (GlcN, D) units, thus yielding chitosans. The presence of GlcN units conveys cationic charges and leads to water solubility at slightly acidic pH values, making chitosans unique in terms of both their physicochemical properties and their biological activities. As a consequence, numerous possible applications have been described for chitosans, including but not limited to a purifying agent for drinking water and wastewater, fat-blocker and functional food additive, antimicrobial agent, stabilizer in paper and textile manufacturing, moisturizer and emulsifier in cosmetics, plant growth promoter and strengthener in agricultural plant disease protection, as a controlled-release drug and gene delivery system, a mucoadhesive factor, and even a hemostatic wound dressing that can promote scar-free wound healing.1–4 Today’s commercially available chitosans are mainly obtained from shrimp or crab shell wastes,5,6 i.e., byproducts of the fishing industry. They are processed using chemical and physical routes, resulting in products varying in composition and quality. Chitosan polymers may differ in their degree of polymerization (DP), their fraction of acetylation (FA), and their pattern of acetylation (PA). These parameters have been shown or proposed to strongly influence the physicochemical properties and therefore the biological activities of chitosans.7,8 Hence, accurate structural characterization is a key factor in understanding structure-function relationships of chitosans. To characterize biological polymers, including chitosans, numerous analytical methods have been described, such as X-ray spectroscopy, Fourier transform infrared (FTIR) spectroscopy,
and nuclear magnetic resonance (NMR) spectroscopy.9 However, these methods require substantial amounts of sample, typically in the multi-milligram range, which is not always available at lab scale or in early stages of research. More recently though, possibilities for biological polymer analysis have been expanded by fingerprinting techniques, typically involving partial depolymerization of the polymers and yielding mixtures of oligomers, which are subsequently analyzed.10 If combined with mass spectrometry (MS), these fingerprinting techniques have the advantage of extremely high sensitivity while at the same time reducing the sample amount from the milligram to the micro- or nanogram range. Fingerprinting techniques require partial depolymerization of the sample, which can be achieved using chemical, physical, or enzymatic10,11 treatments. Enzymatic depolymerization might be preferred because enzymes often have higher cleavage specificities than the other methods. For example, in chemical hydrolysis of heteropolysaccharides consisting of different monosaccharide building blocks, the specificity at best concerns a single residue, e.g., selective acid hydrolysis of rhamnosyl linkages in rhamnogalacturonans or HF solvolysis of the glycosidic linkage of methyl-esterified galacturonic acids in partially methyl-esterified homogalacturonans.12 In contrast, lytic or hydrolytic enzymatic cleavage of pectins using pectin lyases or endopolygalacturonases may require a sequence of two consecutive methyl-esterified or two consecutive free galacturonic acid residues for cleavage,13,14 greatly increasing the analytical power of the fingerprinting technique. Similarly, fingerprinting analyses of partially acetylated chitosans may be achieved using either partial chemical depolymerization or enzymatic hydrolysis. Chemical approaches include selective acid hydrolysis of the glycosidic linkage of GlcNAc residues or selective oxidative deamination of GlcN residues. Enzymatic approaches can potentially involve a wide range of chitosan hydrolyzing enzymes, particularly chitinases
ACS Paragon Plus Environment
Analytical Chemistry
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
and chitosanases. Chitinases are able to cleave the glycosidic linkage between two adjacent GlcNAc units (A/A), and some chitinases may also cleave the GlcNAc-GlcN (A/D) or the GlcN-GlcNAc (D/A) linkage.15 Similarly, chitosanases cleave the glycosidic linkage between two adjacent GlcN units (D/D), and some can also cleave A/D or D/A. Clearly, the more specific enzymes, which selectively cleave only A/A or only D/D, are preferred for fingerprinting purposes. Recently, we described16 another specific chitosan hydrolyzing enzyme, called chitinosanase, which exclusively cleaves after a GlcN followed by a GlcNAc unit (DA/XX). This absolute specificity involving both an acetylated and a deacetylated unit makes this enzyme ideally suited for fingerprinting analyses of chitosan polymers with a wide range of FA. However, analyzing enzymatically produced oligomeric fingerprints by mass spectrometry leads to large and complex data sets. Comparable data obtained in the field of chemometrics, and, more recently, from metabolomic17 or proteomic analyses in biological, medical, or agricultural research, have increasingly been analyzed using multivariate data analysis techniques such as principal component analysis (PCA)18 or partial least squares (PLS) regression19,20. Thus, we apply those techniques in the chitinosanase-based enzymatic / mass spectrometric fingerprinting approach we present here. We determine the FA of microgram amounts of chitosan polymers (having a FA range of 0.10-0.60) by combining UHPLC-ESI-MS, in silico simulations, and multivariate data analysis. 1H-NMR is used to validate the results of our method.
MATERIALS AND METHODS Enzyme production and sample preparation. Chitinosanase was produced as previously described by Kohlhoff et al.16. Chitosan samples of varying FA were produced using fully deacetylated chitosan (FA = 0.0) provided by Gillet-Mahtani chitosan PVT.LTD (Gujarat, India). Chitosan was N-acetylated by acetic anhydride as previously described21, producing FA from 0.10-0.60. FA and Mw were confirmed by NMR and gel permeation chromatography, respectively. Before chitinosanase treatment, 30 µg of the prepared chitosan polymers were dissolved in 30 µl of 200 mM ammonium acetate buffer, pH 4.2. For depolymerization, 2 µl of chitinosanase (100 µg/ml) was added and samples were incubated for 65 h. The reaction was stopped by evaporation (speed vac). To remove residual volatile buffer components, samples were washed three times with 50 µl Milli-Q water. Prior to mass spectrometric analysis, oligomers were dissolved in 30 µl MilliQ water. Nuclear magnetic resonance (NMR) spectroscopy. For each measurement, approximately 5 mg of chitosan were dissolved overnight in 1 ml acidic solution of D2O (1 ml 99.9% D2O and 2 µl DCl) and subsequently lyophilized. The procedure was repeated twice. The FA of chitosan samples was determined using proton NMR as proposed by Hirai et al.22 and Lavertu et al.23. Spectra were recorded using an AV300 or DPX300 300 MHz spectrometer (Bruker, USA) respectively. High performance-gel permeation chromatography (HPGPC). Gel permeation chromatography (3x Novema® column PSS 30000 Å and guard column I.D.: 8mm) was carried out to determine the weight average molecular weight (Mw) of chitosan samples. The column system was coupled online with a refractive index detector (Agilent Serie 1200 RID®) and a
Page 2 of 8
multi-angle laser light scattering photometer (PSS SLD 7000 MALLS®) equipped with a 5 mW He/Ne laser operating at λ = 632 nm. The dn/dc values used for Mw determination were deduced from a polynomial based on previous studies relating the dn/dc with the FA of a sample.24 Samples were prepared in 50 mM ammonium acetate buffer (pH 4.5) in a concentration of 2 mg/ml. Depending on the expected Mw, 50-100 μl of sample was injected and the flow rate was kept constant at 0.6 ml/min. Data were evaluated using WinGPC software by PSS. Based on Mw and FA, corresponding DPs were calculated. UHPLC-ESI-MS and data acquisition. Ultra-high performance liquid chromatography – electrospray ionization – mass spectrometry (UHPLC-ESI-MS) was used to semi-quantitatively determine the amounts of different chitooligosaccharides (COS) derived from enzymatic hydrolysis of chitosans. The COS were separated by hydrophilic interaction liquid chromatography (HILIC) using a Dionex Ultimate 3000RS UHPLC system (Thermo Scientific, Milford, USA) equipped with an Acquity UPLC BEH Amide column (1.7 µm, 2.1 mm x 150 mm; Waters Corporation, Milford, MA, USA) and a VanGuard precolumn (1.7 µm, 2.1 mm x 5 mm; Waters Corporation, Milford, MA, USA). This system was coupled to an amaZon speed ESI-MSn detector (Bruker Daltonik, Bremen, Germany). The applied method was carried out as described by Hamer et al.25 with following modifications: 2 µl of hydrolysate in a concentration of 1 mg/ml were injected into the system by an autosampler. The flow rate was set to 0.4 ml/min and temperature of the column oven was set to 35°C. Separation of COS having DP 2-9 was achieved within 32 min with eluents A (80:20 acetonitrile/water) and B (20:80 acetonitrile/water), both containing 10 mM NH4HCO2 and 0.1% (v/v) HCO2H. The following elution profile was applied: 0-3 min, isocratic 100% (v/v) eluent A; 3-23 min, linear from 0%-85% (v/v) eluent B; 23-28 min, isocratic 85% (v/v) B; followed by column re-equilibration: 28-29 min linear from 85%-0% (v/v) B and then 29-32 min, isocratic 100% (v/v) A. The target mass was set to m/z 700. Raw MS data were converted to mzML file format and spectra were analyzed by Python scripts using pymzML26. Individual steps of the data analysis are described in detail in Supporting Information page S-2. COS with DP 2-9 and every possible FA were analyzed by determining relative MS signal intensities of the predominantly formed [M + H]+ ions with charges 1-3. For COS dimers, sodium adduct ions ([M + Na]+) and dimeric ions ([2M + H]+ and [2M + Na]+) were also observed and therefore included for quantification; they represented up to ~50% of the total DP 2 COS signal. Monoisotopic and second isotopic peak intensities were integrated over elution time of COS and were required to exceed a signal-to-noise ratio (SNR) of three. For further analysis, the sum of relative intensities of all detected COS was normalized to one. These normalized intensities of DP 2-9 COS are referred to as “in vitro fingerprints”. A table with in vitro fingerprints is available in supporting information SI-2. In silico simulation of enzymatic hydrolysis. Chitosan polymers were computationally simulated according to a random distribution of GlcN and GlcNAc residues as in chemically reacetylated chitosan polymers27,28 with the probability of each residue to be acetylated p(GlcNAc) = FA. Simulations included a FA range of 0.044-0.96 (steps of 0.02) and a DP range of 252500 (25-200 in steps of 25; 200-2500 in steps of 100) to cover both the FA and DP range of all chitosans used in vitro.
ACS Paragon Plus Environment
Page 3 of 8
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
Figure 1. Principal component analyses of COS derived from chitosan samples varying in FA. Depicted are (top) score plots and (bottom) loading plots of PCAs carried out on (left) in silico and (right) in vitro fingerprints. PC1 and PC2 explain more than 96% of the variability in fingerprints. PC1 scores separate fingerprints by the substrate FA. Corresponding loadings reveal that separation is correlated with the number of GlcNAc residues in COS.
For each distinct DP and FA, five replicates with 10,000 molecules per replicate were generated. In silico chitosans were cleaved according to the chitinosanase cleavage specificity DA/XX. Oligomers produced by in silico simulation with DP 2-9 are referred to as “in silico fingerprints,” and the sum of all DP 2-9 oligomers within one fingerprint was normalized to one. A table with in silico fingerprints is available in supporting information SI-3. Principal component analysis (PCA). PCA was used to visualize in vitro and in silico enzymatic fingerprints. Only oligomers found in more than 60% of the samples were included. Within each sample, relative intensities of remaining oligomers were summed up and normalized to one. Each data set was mean-centered and scaled to unit variance. Finally, PCAs based on singular value decomposition were performed using the prcomp() function from the stats package in R29. Partial least squares (PLS) regression. PLS was used to create in silico fingerprint based models for FA prediction of chitosans. Fingerprints were normalized to one and values for predictors were mean-centered and scaled to unit variance. PLS regression was carried out using the plsr() function from R package pls30, and the implemented kernel algorithm was used to model relationships between predictor variables (X, COS) and the response variable Y (FA). To find optimal numbers of
PLS components, the amount of variance (in X and Y) explained by the models was determined and root mean square errors (RMSE) of calibration (RMSEC) were calculated. For cross-validation, data sets were divided into five consecutive segments and the bias-corrected RMSE of cross-validation (RMSECV)31 were calculated. For model validation, we used in vitro fingerprints of chitosans with known FA (determined by 1 H-NMR) to calculate RMSE of prediction (RMSEP).
RESULTS AND DISCUSSION Preparation and characterization of chitosan polymers. The aim of this study was to determine if chitinosanase is appropriate for performing enzymatic fingerprinting analyses of polymeric chitosans. To this end, we first generated a series of chitosan polymers with similar average degree of polymerization (determined by HP-GPC-RID-MALLS), but whose fraction of acetylation varied from approximately 0.10-0.60 (determined by 1H-NMR). Chitosans were produced by partial re-Nacetylation of fully de-N-acetylated polyglucosamine polymers (Table 1).
ACS Paragon Plus Environment
Analytical Chemistry
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Page 4 of 8
Figure 2. Partial least squares regression (PLSR) model development. (a) The explained variance and cumulative explained variance in in silico fingerprints (X variables; grey circles) and in the FA (response Y; black diamonds) depending on the number of included PLS components. (b) The root mean square errors (RMSE) of calibration (RMSEC; black squares) and cross-validation (RMSECV; grey triangles) are depicted for PLSR models with different numbers of PLS components.
Table 1. Chitosan polymer series with similar DP and varying FA (batch 1). Sample
FA
Mw [Da]
DP
1
0.16
220,000
1312
2
0.22
170,000
999
3
0.35
210,000
1197
4
0.50
221,000
1214
5
0.60
242,000
1300
FA values were determined via proton NMR.
Enzymatic / mass spectrometric fingerprinting. Next, chitosans were treated with chitinosanase until complete enzymatic hydrolysis. Base peak chromatograms of all samples are shown in Figures S-2, S-3, and S-4. Example spectra of COS are shown in Figure S-5. A complete hydrolysis was verified by measuring oligomeric fingerprint products via UHPLC-ESIMS at different time points of incubation until stable compositions were measured (data not shown). To estimate the accuracy and coverage of product analysis, all in vitro experiments were additionally simulated in silico. Therefore, chitosans with the same FA and DP as those used in in vitro experiments were simulated and hydrolyzed according to the cleavage specificity of chitinosanase DA/XX. Resulting in silico fingerprints were compared to their corresponding in vitro ones. Although the amounts of oligomers in in vitro fingerprints were determined in a semi-quantitative manner, in vitro and in silico fingerprints turned out to be highly similar (data not shown). This not only confirmed that our experimental analysis did cover the majority of oligomeric products, but it also indicates – by what we call
“reverse fingerprinting” – the previously reported absolute cleavage specificity of chitinosanase for the DA/XX motif. For further experiments, we decided to restrict fingerprints to DP 29 oligomers, since they were easily detectable with UHPLCESI-MS in all fingerprints of the FA series, and, according to in silico simulation, higher-DP oligomers rarely occur in fingerprints of intermediate FA. Principal component analysis. To visualize in vitro (batch 1) and in silico (FA 0.16, 0.22, 0.35, 0.50, and 0.60; DP 8001100) fingerprints, we performed separate principal component analyses (PCA) (Figure 1). In both cases, score plots (Figure 1 scores) revealed that all chitosan samples could be unequivocally separated using their scores for the first two PCs. In sum, PC1 and PC2 explained more than 96% of the variance in both the in silico and in vitro data; where PC1 scores separate chitosans solely according to their FA. Corresponding loadings of COS for the first two PCs (Figure 1 loadings) show that the FA of the separated chitosans is also represented by the variables, i.e., oligomers, responsible for separation. As previously hypothesized, this shows that PCA is well suited to discriminate oligomer mixtures that are produced by enzymatic hydrolysis of chitosan substrates varying in FA. Moreover, it led us to the conclusion that some information about the FA is already contained in PC1, and, as a consequence, that chitinosanase fingerprints should be appropriate for creating predictive models that can help determine the FA of unknown chitosan samples. Partial least squares regression, model development and validation. In a next step, we developed a partial least squares regression (PLSR) model to predict the FA of chemically re-Nacetylated chitosan samples (having unknown FA). To maximize the predictive power of the model, underlying data should
ACS Paragon Plus Environment
Page 5 of 8
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Analytical Chemistry
cover a wide range of substrates, both in terms of FA and DP. Therefore, because the number of in vitro fingerprints that can be produced is limited, and because in silico simulations on chitinosanase generated data highly similar to in vitro results, we decided to base the model exclusively on in silico data. This not only allowed us to include an unlimited number of possible fingerprints, but it also avoids measurement errors associated with NMR or MS analyses. We included chitosans with FA in a range from 0.04-0.96 and DP in a range from 25-2500. After the simulated chitinosanase treatment, all 52 occurring DP 2-9 oligomers (X variables) were used as predictors for the FA (response, Y). Each fingerprint was normalized to one, and predictor variables were mean-centered and scaled to unit variance. To determine the optimal number of PLS components to be included in the final model, we first evaluated individual and cumulative variance in X and Y explained by the components as well as the root mean square errors of calibration (RMSEC) and cross-validation (RMSECV) (Figure 2). Figure 2a shows that six components are needed to explain almost 100% of the variance in X, whereas only two components are needed to reach a comparable value for Y. RMSEC were calculated for different PLSR models with 1-20 PLS components. Figure 2b reveals that the RMSEC is
