i An update to this article is included at the end

Carbohydrate Polymers 212 (2019) 186–196

Contents lists available at ScienceDirect

Carbohydrate Polymers
journal homepage: www.elsevier.com/locate/carbpol

ATR-FTIR spectroscopy to determine cell wall composition: Application on a T

large diversity of fruits and vegetables
⁎
Maria H.G. Canteria, , Catherine M.G.C. Renardb, Carine Le Bourvellecb, Sylvie Bureaub
a
Department of Chemistry and Biology-DAQBI, Federal University of Technology–Paraná, Francisco Beltrão, Brazil
b
UMR408 SQPOV, Sécurité et Qualité des Produits d’Origine Végétale, INRA, Avignon University, F-84000 Avignon, France

A R T I C LE I N FO A B S T R A C T

Keywords: Infrared spectroscopy coupled with multivariate analyses such as linear regressions was applied to assess the
Attenuated Total Reflectance Fourier main cell wall components of a huge diversity of fruits and vegetables belonging to 29 plant species. The
Transform Infrared methodology was tested on the raw freeze-dried powders and on their corresponding AIS (Alcohol Insoluble
Alcohol insoluble solids Solids) dried by solvent exchanges. The most informative spectral region was 1750-1035 cm−1. Excellent pre-
Pectins
dictions (determination coefficient R2 ≥ 0.9 and residual predictive deviation RPD ≥ 3.0) were obtained for AIS
Cellulose
Partial Least Square Regression
yields and for arabinose, total glucose, non-cellulosic glucose, total neutral sugars, methanol and starch contents
in the AIS samples. The key wavenumbers were: 1740 cm-1 for total neutral sugars; 1075, 1440–1450, 1616 and
1740 cm-1 for pectins; 895, 1035–1041 and 1160-1163 cm-1 for cellulose and 1035–1041 cm-1 for lignin.
Limitations of the reference methods to analyze cell wall components (biochemical assays, spectrophotometry,
chromatography) affecting the prediction accuracy were also discussed.

1. Introduction
The fast quantification of plant cell walls and their components
remains a challenge today. It is needed to better quantify dietary fiber
intake, as cell walls are the main source of dietary fiber (Niklas, 2004)
and to understand the functionality of many plant foods. The analytical
cell wall characterization is time consuming and demands a high level
of expertise, while the cell wall composition is diverse and highly
variable. There are differences between species and individual plant
organs, with specific components found inside particular lineages, tis-
sues, or cells, or still controlled temporally or spatially at the single cell
level, depending on the developmental stage and season (Popper et al.,
2011).
Cell walls can be described as a network of cellulose micro fibrils
embedded in an amorphous matrix composed mainly of pectins and
hemicelluloses (Carpita & Gibeaut, 1993). Non-lignified and highly
hydrated primary cell walls prevail in plant foods for Human
consumption. The primary cell walls of higher plants contain mostly
polysaccharides such as cellulose, hemicelluloses, or pectins, and minor
amounts of structural glycoproteins, phenolic compounds, bound mi-
nerals and enzymes, with both qualitative and quantitative variations
(Costa & Plazanet, 2016; Fahey, Nieuwoudt, & Harris, 2017;
Szymanska-Chargot, Chylinska & Zdunek, 2015). In general, conven-
tional methods can be used to evaluate the composition, structure and
organization of cell wall polymers and spectral methods to screen a
large cell wall population for industrial applications (Costa & Plazanet,
2016; Popper et al., 2011; Sørensen, Domozych, & Willats, 2010). In
both methods, the acid by-products must be removed before analysis
and many manual processing steps are needed (Ruiz-Matute,
Hernandez-Hernandez, Rodriguez-Sanchez, Sanz, & Martinez-Castro,
2011). Even in our laboratory, which has a validated methodology,
specific equipment and specialized technical staff, the average analysis
time is 24 h for seven samples simultaneously characterized. Due to the
time and the manual work necessary for their characterization, the need

Abbreviations: AIS, alcohol insoluble solids; Ara, arabinose; ATR-FTIR, Attenuated Total Reflectance Fourier Transform Infrared Spectroscopy; AUA, anhydrous
uronic acid; DM, degree of methylation; DW, dry weight; Fuc, fucose; Gal, galactose; Glc, glucose; Lig, lignin; LV, latent variables; MeOH, methanol; MVA,
Multivariate analysis; NCGlc, non-cellulosic glucose; PCA, Principal Component Analysis; PLS, Partial Least-Squares; R2, coefficient of determination for the vali-
dation set; Rha, rhamnose; RM, raw material; RMSECV, root mean square error of cross-validation; RPD, residual predictive deviation; SD, standard deviation; SNV,
standard normal variate; Xyl, xylose
⁎
Corresponding author at: Federal University of Technology, Department of Chemistry and Biology-DAQBI, Francisco Beltrão, Linha Santa Bárbara, s/n, Francisco
Beltrão, Paraná, Brazil.
E-mail addresses: canteri@utfpr.edu.br, canteri.mhg@gmail.com (M.H.G. Canteri), catherine.renard@inra.fr (C.M.G.C. Renard),
carine.le-bourvellec@inra.fr (C. Le Bourvellec), sylvie.bureau@inra.fr (S. Bureau).

https://doi.org/10.1016/j.carbpol.2019.02.021
Received 21 December 2018; Received in revised form 6 February 2019; Accepted 7 February 2019
Available online 08 February 2019
0144-8617/ © 2019 Elsevier Ltd. All rights reserved.
M.H.G. Canteri, et al.
to develop alternative methods such as ATR-FTIR to predict the cell
wall neutral sugars is reinforced.
The advantage of the Attenuated Total Reflectance Fourier
Transform Infrared spectroscopy (ATR- FTIR) is to generate spectra on a
wide range of samples (powders, liquids, pastes) with a minimum of
sample preparation, reducing the analysis time (Bureau, Ścibisz, Le
Bourvellec, & Renard, 2012; Fahey et al., 2017; Fellah, Anjukandi,
Waterland, & Williams, 2009; López-Sánchez, Ayora-Cañada, & Molina-
Díaz, 2009; Szymanska-Chargot & Zdunek, 2013). Moreover, to extract
and exploit the information contained in spectral data, multivariate
statistical techniques are required and different methods are used with
two objectives: 1) quantitative applications based on the relationship
between spectral and reference data obtained by the conventional
methods in order to establish predictive models and 2) qualitative ap-
plications to illustrate the studied diversity and classify samples ac-
cording to their spectral characteristics. However, significant data, co-
herent methods and a great comprehension of the analysis purpose are
important to reach significant results (Lee, Liong, & Jemain, 2017;
Nunes, Alvarenga, de Souza Sant’Ana, Santos, & Granato, 2015).
Since the 1970s, ATR-FTIR has been used for fast cell wall char-
acterization. The primary focus was on the analysis of pectins and their
degree of esterification (Barros et al., 2002; Bichara et al., 2016; Bociek
& Welti, 1975; Chatjigakis et al., 1998; Coimbra, Barros, Barros,
Rutledge, & Delgadillo, 1998; Ferreira, Barros, Coimbra, & Delgadillo,
2001; Filippov & Chernei, 2002; Filippov & Kohn, 1975; Filippov, 1992;
Kyomugasho, Christiaens; Shpigelman, Van Loey & Hendrickx, 2015;
Chylińska, Szymańska-Chargot, & Zdunek, 2016). This is due to the
existence of two distinct and fairly isolated peaks representing respec-
tively the methyl-esterified and the non-methyl-esterified galacturonic
acid moieties in the 1750 – 1600 cm−1 area, and the difficulties to
correctly assess these components with conventional method. The of-
ficial industrial method (by titrimetry) is time-consuming, requires
about 1 g of pectins, and demands care in avoiding carbonatation of the
pectin solution and titrating alkali. In literature, the most reliable
methods are still delicate and dangerous colorimetric assays using
concentrated sulfuric acid (Blumenkrantz & Asboe-Hansen), as quanti-
tative hydrolysis or methanolysis of pectins is not feasible. Later, more
advances in ATR-FTIR technics and chemometrics have allowed the
determination of cell wall polysaccharides (Robic, Bertrand, Sassi,
Lerat, & Lahaye, 2009), especially of cellulose and hemicelluloses
(Abidi, Cabrales, & Haigler, 2014; Garside & Wyeth, 2003; McCann
et al., 2007), lignin (Faix, 1991; Tamaki & Mazza, 2011; Toribio-Cuaya
et al., 2014), and starch (Demiate, Dupuy, Huvenne, Cereda, &
Wosiacki, 2000; Kizil, Irudayaraj, & Seetharaman, 2002; Sevenou, Hill,
Farhat, & Mitchell, 2002). However, in most of these studies, a limited
set of plant samples with a slight variability was used. Our previous
works have been dedicated to the use of ATR-FTIR coupled with mul-
tivariate analyses such as PLS regression and PCA to assess the com-
position of sugars, organic acids and polyphenols in different fruit
species and specially on apples using dry samples (Bureau et al., 2012)
with a particular attention done on the sampling which should be re-
presentative of the observed and known variability. In this work, the
objective was to use the same methodology and strategy to evaluate the
most as possible all information on cell wall, quantity and composition,
taking into account the fragmented cell wall characterization described
in literature above.
Our hypothesis is that ATR-FTIR coupled with chemometrics can be
a generic tool to characterize cell walls, both on their abundance and on
their composition, taking into account a large diversity of plants. To test
it, the yield and the composition of cell walls were investigated in a
large range of fruit and vegetables using both, spectral data and con-
ventional methods. To evaluate the available information according to
the sample preparation, two types of homogeneous samples, freeze-
dried raw material and alcohol insoluble solid materials, were scanned
in ATR-FTIR spectroscopy. Discrimination of samples by PCA (Principal
Component Analysis) and prediction of cell wall composition using
187
Carbohydrate Polymers 212 (2019) 186–196
linear regression (PLS, Partial least squares) were then attempted.
2. Material and methods
2.1. Chemicals
Ammonium oxalate, methanol and phenol were from Merck
(Darmstadt, Germany). Acetic acid, acetic anhydride, anhydrous so-
dium carbonate, ethanol-d3, Folin and Ciocalteu’s phenol reagent,
formic acid, galacturonic acid, gallic acid, glucose, hydrochloric acid,
inositol, lignin alkali, N-methylimidazole sodium acetate, sodium bor-
ohydride (NaBH4) and sodium hydroxide were from Sigma-Aldrich
(Deisenhofen, Germany). Sugars (arabinose, fucose, galactose, man-
nose, rhamnose and xylose) were from Fluka (Buchs, Switzerland).
Acetone and ethanol and were from Fisher Scientific (Strasbourg,
France).
2.2. Raw material (RM) and alcohol insoluble solids (AIS) extraction
Fresh fruit and vegetable samples from France were purchased at
local grocery and immediately manually peeled and cut in small pieces
(˜3 cm3). All of these samples were freeze-dried and ground before
storage in hermetically closed flasks. The fruits and vegetables from
Brazil, obtained directly from the orchards, were dehydrated (50 °C
until constant weight). In order to increase the phenotypic variability,
these samples were harvested at several ripening stages and some of
them were submitted to a thermal inactivation. All these dehydrated
powders were designated as raw material (RM). Alcohol Insoluble
Solids (AIS) from these RM were isolated using the method described by
Renard (2005). The codes and binomial nomenclature as well as some
processing characteristics of these samples are presented in Table 1.
2.3. Characterization of AIS by biochemical analyses
All cell wall characterizations were performed on the AIS samples.
Neutral sugars (as alditol acetates after hydrolysis and derivatisation by
gas chromatography -GC), anhydrous uronic acid-AUA (by colorimetry
using the meta-hydroxyl-diphenyl assay), methanol-MeOH (by stable
isotope dilution assay by headspace-GC–MS after saponification) were
analyzed as described by Renard and Ginies (2009). Degree of methy-
lation (DM) was calculated as the molar ratio of methanol to uronic
acids. In order to determine cellulosic glucose, neutral sugars included
NCGlc and cellulosic Glc were analysed as alditol acetates after acid
hydrolysis with two options. For cellulose analysis, samples were sub-
mitted to 13 mol/L sulfuric acid prehydrolysis. To characterize NCGlc
no prehydrolysis was carried out, and cellulosic Glc was calculated as
the difference between glucose contents after prehydrolysis by 13 mol/
L sulfuric acid and without prehydrolysis. Starch was quantified in
samples positive for the Lugol iodine test using the Megazyme total
starch kit (α-amylase K-TSTA 05/2008; Megazyme International Ire-
land Ltd, Ireland, UK). Total polyphenol content was estimated by the
Folin-Ciocalteu method (Singleton & Rossi, 1965) adapted to micro-
plates, using gallic acid as a standard. Polyphenols were extracted from
50 mg of AIS by 0.7 mL of acidified methanol (10 mL acetic acid per L)
for 15 min in an ultrasonic bath (Bioblock Scientific - Fisher Scientific,
Illkirch, France), centrifuged and filtered (0.45 μm). For analysis, 25 μL
of Folin's reagent and 20 μL of sample or standard were deposited in a
96-well microplates (Greiner Bio-one, VWR, France). Subsequently,
200 μL of Na2CO3 (0.4 mol L−1) were added using automatic injectors
connected to the spectrofluorimeter (Safas, Monaco). The absorbance
was read at 730 nm after 20 min stirring. Lignin was quantified as de-
scribed in Brahem, Renard, Gouble, Bureau, and Le Bourvellec, (2017).
2.4. ATR-FTIR spectra
Both, RM and AIS, were passed through a 150 μm (100 Mesh) size
M.H.G. Canteri, et al. Carbohydrate Polymers 212 (2019) 186–196

Table 1 For each cell wall parameter, analysis of the ATR-FTIR data involved 10
Identification of the Brazilian and French fruit and vegetable samples used as separate modelling/validation exercises and two spectral areas were
raw material to study their cell wall composition by both biochemical char- tested (4000-700 cm−1 and 2000-800 cm−1). The absorption band
acterizations from AIS and ATR-FTIR. around 2400 and 2200 cm−1 was discarded prior the calculation. The
Specie Origin Binomial nomenclature Code(s) SNV (standard normal variate) correction was applied to each spec-
trum. Partial least-squares (PLS) and more exactly PLS1 (only one
Apple* France Malus domestica APPf and
variable at a time) regression method was performed. The performance
APPp
Banana* France/ Musa sp. BANf and
of the developed models is given in tables as the mean of the 10 in-
Brazil BANp dependent tests, taking into account for each test the highest R2 (de-
Brazilian cherry Brazil Eugenia uniflora PITA termination coefficient of validation) and the lowest RMSECV (root-
Broccoli France Brassica oleracea var. BROC mean-square error of cross-validation) with the number of latent vari-
italica
ables up to a maximum of 15. The residual predictive deviation (RPD)
Cauliflower France Brassica oleracea var. CAUL
botrytis was calculated as the ratio of the standard deviation of the response
Carrot France Daucus carota CARO variable to the RMSECV. The RPD is used to test the accuracy of the
Celery France Apium graveolens CELE developed models: lower than 1.5 is considered insufficient for most
Eggplant Brazil Solanum melongena EGGP
applications, between 2 and 2.5 indicates a possible coarse quantitative
Gabiroba Brazil Campomanesia xanthocarpa GUAB
Guava Brazil Psidium guajava GUAV
prediction, and between 2.5 and 3 or above a respectively good and
Jabuticaba Brazil Plinia cauliflora JABU excellent prediction (Nicolai et al., 2007; Robic et al., 2009).
Leek France Allium porrum L. LEEK
Lemon France Citrus limon Burm. f. CITR 3. Results and discussion
Lettuce France Lactuca sativa SALA
Orange France Citrus sinensis ORAN
Papaya Brazil Carica papaya PAPA 3.1. AIS composition by reference methods
Passion fruit Brazil Passiflora flavicarpa PASS
Peach* France Prunus persica PEAf and The samples were chosen to represent a large diversity of plant
PEAp materials in order to cover the variability of cell wall composition,
Persimmon Brazil Diospyros kaki KAKI
Pineapple France Ananas comosus ANAN
qualitatively and quantitatively. According to Xu, Yu, Tesso, Dowell, &
Pomelo France Citrus maxima PAMP Wang (2013), to develop some generic models of prediction, all para-
Potato* France Solanum tuberosum POTA meters affecting sample characteristics should be considered. In total,
Red black mulberry Brazil Rubus sellowii BBER raw materials (RM) included 60 samples of fruits and vegetables were
Red cattley guava Brazil Psidium cattleianum ARAC
prepared in two countries (Brazil and France), belonging to 29 species
Spinach France Spinacia oleracea SPIN
Squash France Cucurbita moschata PUMP from 15 different botanical orders, with the characterization of four
Sweet potato Brazil Ipomoea batatas PATA tissues (flesh, peel, leaf and tuber) and different ripening stages.
Tomato* France Solanum lycopersicum TOMA The yield of AIS from RM, giving an evaluation of the cell wall
Yerba mate Brazil Ilex paraguariensis HERB quantity, exhibited a large range of variation due to the large diversity
of studied plant materials (11–162 mg g−1 of fresh weight or
Used abbreviations: f-flesh a; p-peel; *indicates that more than one variety has
been evaluated to these samples. 40–259 mg g−1 of dry weight). Large ranges were observed too for the
measured cell wall components (Supplementary Table 1) which were
screen sieve and stored in P2O5 atmosphere before acquisition of ATR- illustrated in the Fig. 1. All the studied parameters were statistically
FTIR spectral data (4000 cm−1 to 650 cm−1) using the Tensor 27 FTIR different according to the large number of species and tissues, except
spectrometer (Bruker Optics) equipped with a single-reflectance hor- for the NCGlc between tissues, for which the analytical uncertainty is
izontal ATR cell (Golden Gate equipped with a diamond crystal, Bruker very high compared to the measured content (Supplementary Table 1).
Optics), as described by Bureau et al. (2012). Each sample was mea- The AIS yield and composition were not compared and discussed in this
sured five times (by removing and putting different aliquots of powder paper.
to evaluate their heterogeneity) and each spectrum was the average of Some studies on the AIS composition of different plants are in ac-
16 scans. In total, 60 raw samples (RM) and 60 AIS samples were cordance with our data (Caffall & Mohnen, 2009; Houben, Jolie,
scanned with five replications. Fraeye, Van Loey, & Hendrickx, 2011; Kosmala et al., 2013; Kurz,
Leitenberger, Carle, & Schieber, 2010; Renard, Voragen, Thibault, &
Pilnik, 1990; Szymanska-Chargot et al., 2015; Szymańska-Chargot,
2.5. Statistics Chylińska, Gdula, Kozioł, & Zdunek, 2017). Starch was usually very low
as most samples do not accumulate reserves (leaves) or do not have any
All biochemical analyses were performed at least in triplicate on more at a ripe stage. But, some samples such as banana, sweet potato
different aliquots of each powdered sample. These results were in- and potato always remained very rich in starch.
dicated as mean values of replicates and the reproducibility of the re- The variability in AIS samples composition is illustrated (Fig. 2) by
sults was expressed as the pooled standard deviations using the sum of the 2-D plane defined by the two first principal components (PC1 and
individual variances pondered by the individual degrees of freedom PC2) of a PCA performed on the contents of AIS components. PC1 and
(Box, Hunter, & Hunter, 1978; Kosmala et al., 2013). PC2 explained respectively 39.2% and 19.3% of the total variance. PC1
A one-way ANOVA was performed on physicochemical parameters was highly positively correlated to the main neutral sugars (Rha, Ara,
using the Statistica for Windows 5.0 (StatSoft, Tulsa, Oklahoma, USA) Fuc, Man, Gal) and the pectic fraction characteristics (AUA, MeOH and
to evaluate the effect of tissues and species on the studied variables DM) and negatively correlated to the starch, NCGlc and AIS yield. PC2
(except for AIS yield). was positively correlated to starch and NCGlc and strongly negatively
Spectral preprocessing and multivariate data analysis were per- correlated to Lig and Xyl content. The sample map presented two dis-
formed with Matlab 7.5 (Mathworks Inc., Natick, MA) software using tinct groups. The samples including potato (POTA), banana (BANA) and
the SAISIR package (SAISIR, free procedures using MATLAB for che- sweet potato (PATA) with a high starch content were separated from all
mometrics, available at http://easy-chemometrics.fr. The samples were others. A few samples were intermediate and close to the center: ja-
randomly distributed among the calibration (9/10 giving 270 data) and boticaba (JABU), yerba mate (HERB) and banana peel (BANp). In the
prediction (1/10 giving 30 data) sets to perform a cross-validation test. cloud of low-starch samples, at the opposite side, apple flesh (APPf),

188
M.H.G. Canteri, et al. Carbohydrate Polymers 212 (2019) 186–196

Fig. 1. Variability of cell wall components of selected

fruits and vegetables (mg g−1).
Abbreviations used: Ara—arabinose, AUA—anhydrous
uronic acid, CGlc—cellulosic glucose, Fuc—fucose,
Gal—galactose, Lig—lignin; Man—mannose,
NCGlc—non cellulosic glucose with correction of the
starch content (glucose from simple hydrolysis),
Rha—rhamnose, STA- starch, Xyl-xylose, *outliers
(Rha=2 lower; Man=2 upper; Xyl=6 upper; Gal=2
upper; NCGLC= 3 upper; Sta= 6 upper).

carrot (CARO), peach flesh (PEAf), orange (ORAN), lemon (CITR) and
Brazilian cherry (PITA), rich in pectins and neutral sugars (except Xyl)
were distinguished from the Brazilian native berries species, relatively
rich in Xyl and Lig, such as red cattley guava (ARAC), guava (GUAV),
gabiroba (GUAB), red black mulberry (BBER) and persimonn-Guiombo
variety (KAKI). These Brazilian fruits present higher crunchiness and
firmness (sensorial data not published), associated to a high mechanical
strength due to the lignin content and a high rigidity due to hemi-
celluloses content (Hayashi & Kaida, 2011; Liu, Luo, & Zheng, 2018).
They also probably contain secondary tissues, since these smaller fruits
were totally ground, without a separation of their different fractions
(peel, flesh and seeds).

3.2. ATR-FTIR spectroscopy

ATR-FTIR spectra were performed directly on the raw freeze-dried

samples and on the AIS samples. According to Türker-Kaya and Huck
(2017), typical ATR-FTIR spectra show several absorbance peaks due to
fundamental transitions and distributed in four discriminated regions:
the XeH stretching region (4000–2500 cm−1), the triple-bond region
(2500–2000 cm−1), the double-bond region (2000–1500 cm−1) and the
fingerprint region (1500–600 cm−1). However, the region from
1800 cm-1 to 1400 cm−1 provides information about the functional
groups occurring in our studied molecules (Szymanska-Chargot &
Zdunek, 2013). Due to the complexity of the cell wall composition, it is
not always possible to assign exactly each ATR-FTIR band to its re-
spective functional chemical group or compound. The variation of
spectra with distinction or lack of some bands illustrated the cell wall
variability. The observation of the spectral data obtained on the sixty
samples of fruits and vegetables allowed to highlight the RM and AIS
variability as well as to identify the main absorption bands character-
istics of their composition (Fig. 3).
In these spectra the following peaks, related to cell wall components
(Supplementary Table 2) were commonly found in the two types of
samples (RM and AIS): 3334 and 2917 cm−1 are assigned to cellulose
(Abidi et al., 2014; Talari, Martinez, Movasaghi, Rehman, & Rehman,
2017; Toribio-Cuaya et al., 2014), 1734 cm−1 to pectins with ester
(Alonso-Simón et al., 2011; Fahey et al., 2017; Filippov & Chernei,
2002; Huck, 2015; Monti et al., 2013; Szymanska-Chargot & Zdunek,
2013), 1626 cm-1 to free carboxyl groups (Abidi et al., 2014; Alonso-
Simón et al., 2011; Filippov & Chernei, 2002; Szymanska-Chargot &
Zdunek, 2013; Talari et al., 2017), 1234 cm−1 to proteins (Huck, 2015;
Fig. 2. Principal component analysis (PCA) based on biochemical reference
Talari et al., 2017) and 1015 cm-1 to polysaccharides, sugars and pec-
analyses of AIS from different tissues from 60 samples. Loadings (top) and score
tins (Talari et al., 2017). The bands found only in the RM spectra were:
plot (bottom) were given for the PC1 (A1) and PC2 (A2).
1410 cm-1 (Fahey et al., 2017; Kizil et al., 2002) and 921 cm-1 that are

189
M.H.G. Canteri, et al.
Fig. 3. ATR FTIR spectra in the 4000-650 cm−1 region: mean of spectra of AIS
and RM from sixty fruits and vegetables samples.
assigned to Lig (Fahey et al., 2017) and 770 cm-1 that are assigned to
phenyl groups (Coates, 2000). Finally, the bands featured exclusively in
AIS spectra, visible only after the removal of soluble components such
as proteins, lipids, acids, free sugars and vacuolar phenolics (that re-
mained fixed on the walls during the preparation of the AIS samples)
were: 1520 cm-1 attributed to Lig and phenolic backbone (Alonso-
Simón et al., 2011; Huck, 2015; Palacio et al., 2014), 1440 cm-1 to
pectins (Monti et al., 2013; Talari et al., 2017), 1314 and 1371 cm-1 to
cellulose and xyloglucan (Abidi et al., 2014; Alonso-Simón et al., 2011;
Fahey et al., 2017; Toribio-Cuaya et al., 2014).
A principal component analysis (PCA) was carried out using the
spectral data (Fig. 4) to highlight the distribution of AIS samples.
PC1 and PC2 explained respectively 52.4% and 25.4% of the total
variance. Concerning different samples of different varieties belonging
to the same plant species, a similarity was observed between the re-
spective spectral data, in accordance with the biochemical data not
statistically different (p < 0.05) (data not shown). The region close to
992 cm−1 corresponds to the polysaccharide absorptions
(Supplementary Table 3) with a negative correlation to PC1. The PC1
separated the starch-rich samples on the negative side (on the left side)
from the others (on the right). The regions between 1230–1236,
1518–1521 and 1600-1625 cm−1 were attributed to respectively
hemicelluloses or lignin, lignin and free carbonyl group of pectins with
a positive correlation to PC1. Thus, on the left, the species such as
potato (POTA), banana flesh (BANf) and sweet potato (PATA) were the
richest in starch, in accordance with the results observed on PCA per-
formed on biochemical reference data of AIS samples (Fig. 2). On the
other hand, on the right, spinach (SPIN) was the poorest in starch but it
was relatively rich in lignin, as well as species such as red black mul-
berry (BBER), papaya (PAPA), cauliflower (CAUL) and yerba mate
(HERB) considered relatively rich in hemicelluloses and lignin. The PC2
was highly positively associated to polysaccharides (988 cm−1) and
negatively linked to arabinose/cellulose at 1061 cm−1 and esterified
carboxyl groups of pectins at 1747 cm−1 (Supplementary Table 3). The
samples were then discriminated along the axis 2 which separated the
samples the richest in starch or cellulose (at the top), as potato (POTA),
190
Carbohydrate Polymers 212 (2019) 186–196
banana flesh (BANf), banana peel (BANp), sweet potato (PATA), from
those rich in high methoxyl pectins associated with relatively high
contents of arabinose and cellulose (at the bottom) as apple flesh (APPf)
and peach flesh (PEAf), and those rich in cellulose as squash (PUMP)
and pineapple (ANAN).
In brief, the PCAs obtained on the reference data and the one on the
spectral data both discriminated samples according to their starch
content. ATR-FTIR could be considered as a good, rapid and easy
method to have an overview on the cell wall variability and their
general composition. However, in order to develop some models of
prediction to quantitatively determine the cell wall composition, both,
spectra and reference methods were used on a same set of samples re-
presentative of the expected variability.
In a first instance, PLS was performed on spectra collected on the
RM freeze-dried powder (Table 2). Concerning the cell wall composi-
tion, the prediction of components directly predicted from the freeze-
dried powders were given with R2 from 73% to 83% of the variation in
the data set. The RPD was comprised between 2 and 2.3 and only for
AUA the RPD value was below 2. A possible coarse quantitative pre-
diction can be obtained for these compounds by ATR-FTIR, considering
to have an estimation of the cell wall composition directly on dried raw
samples without preparing AIS materials, saving time and reducing the
use of solvents. However, regarding the values of RMSECV expressed in
percentage (not shown), the errors of prediction were relatively high
and so, their prediction cannot be used in routine. The main results
obtained on RM was the good prediction of the AIS yield.
The prediction of yield of AIS in dry weight gave R2 of 0.92,
RMSCEV of 65 mg g−1 for the cross-validation sets and RPD of 3.5 in-
dicated an excellent prediction. The prediction of the yield of AIS di-
rectly on freeze-dried raw material presents a great interest from a point
of view of Human nutrition. Total dietary fiber (TDF) is composed
mainly of non-starch polysaccharides (cellulose, hemicelluloses and
pectic substances), undigested cell wall proteins, lignin, phenolic
compounds, phytates, resistant starches and other minor compounds
not digestible by secretions of the gastrointestinal digestive tract
(Jones, 2014). In the same way, a very good prediction is found for dry
matter in tomato (340 samples, 39 cultivars) and in apple (168 samples,
7 cultivars, using ATR-FTIR (ATR ZnSe 6 reflections) associated to PLS
cross-validation (Bureau et al., 2012; Ścibisz et al., 2011). However, to
our knowledge, no paper has related the use of FTIR associated with
multivariate analysis and calibration models on fruit and vegetables to
predict the AIS yield directly on the freeze-dried samples.
In a second instance, PLS was performed on the spectra of the AIS
samples. The quality of the prediction was better on AIS than on raw
samples (RM described above) for starch, total neutral sugars, AUA,
MeOH, phenolic compounds, Lig and DM (Table 2). The best predic-
tions were obtained for starch (R2 of 0.96), total glucose (R2 of 0.95),
methanol (R2 of 0.94), arabinose (R2 of 0.91) and total neutral sugars
(R2 of 0.90). The RPD for these components were above 3 or 4 showing
that their prediction was good or excellent. The better prediction ob-
tained for AIS compared to RM could be due to the fact that the main
contaminating compounds affecting the spectral data, such as low
molecular weight sugars (mainly), organic acids, essential oils, free
amino acids, glycosides and proteins were removed in AIS. The spectral
absorption bands were in this case directly due to the cell wall com-
ponents remaining in the AIS samples. Good or excellent predictive
correlation has already been observed for carbohydrate content and Lig
by ATR-FTIR and PLS regression in woody and herbaceous biomass,
corn stover and rice straw, however the preparation of AIS sample was
not indicated (Tamaki & Mazza, 2011; Xu, Yu, Tesso, Dowell, & Wang,
2013). Applications using ATR-FTIR have also been developed to study
gelation and retrogradation of starch with characteristic bands linked to
its molecular conformation in maize, wheat, starch films, potato starch
gel, cassava sour starch and plant roots (Demiate et al., 2000;
Kačuráková & Wilson, 2001; Warren, Gidley, & Flanagan, 2016). An
ATR-FTIR method giving an excellent correlation with the reference
M.H.G. Canteri, et al.
Fig. 4. Principal component analysis (PCA) using FTIR spectra of 60 AIS samples extracted from fruit and vegetables in the range of 4000-650 cm−1 with Scatter of
loading profile of components PC1 and PC2 (a) and scores (b).
enzymatic analysis (R2 of 0.994) is used to measure the remaining
starch of different sources after the fermentation (Udén, 2009).
Szymanska-Chargot et al. (2015) have also shown that in apple, the
prediction of the percentage in the cell wall material is good for
hemicelluloses (R2 of 0.94) and cellulose (R2 of 0.91) even if the pre-
diction of hemicelluloses should be confirmed because Q2 value, an
indicator of future prediction quality, is low.
For the other cell wall components, R2 varied from 0.78 for phenolic
compounds and rhamnose to 0.87 for Xyl and Lig. For some of them
(Xyl, AUA, Lig and DM), RPD were higher than 2.5 making possible
their prediction by ATR-FTIR. For the others, with a RPD lower than
2.5, their prediction can only be considered as semi-quantitative. ATR-
FTIR is not a sensitive method and did not allow the prediction of fu-
cose, mannose, and phenolic compounds (all three being present in low
amounts, respectively an average of 2.6, 14.7 and 0.5 mg g−1). For
galactose, the RPD of 2.4 might be due to its presence in very diverse
forms in cell walls, such as type I or type II arabinogalactans (respec-
tively b-(1–4) and mixed b-(1–3) and (1–6) glycosidic bonds), ga-
lactomanans or galactoxyloglucans (mostly monomeric side chains)
leading to different main absorption bands. The good prediction of
arabinose and rhamnose could be justified by their relatively high
amounts as they enter respectively in the composition of the side chains
and backbone of pectic rhamnogalacturonan I (Voragen, Coenen,
Verhoef, & Schols, 2009). On the other hand, xylose is present in a large
proportion in xylan chains with varied substitutions or xyloglucans as
non-reducing terminal sugar or within a longer side chain
(Amarasekara, 2013) or as a subunit in xylogalacturunan chains of
pectins (Schols, Bakx, Schipper, & Voragen, 1995).
191
Carbohydrate Polymers 212 (2019) 186–196
Better results are obtained to predict the xylose content of the
hemicellulose polysaccharides (R2 of 0.96 and RMSEP of 10.7%) in
olive pulp (Coimbra et al., 1998) and to predict the uronic acid content
of pectic fractions of olive pulp (R2 of 0.989 and RMSEP of 2.01%) and
of orange pulp (R2 of 0.972 and RMSEP of 2.38%) (Coimbra et al.,
1998). However, in these previous works, the calibration is developed
on a set of non-independent fractions of cell wall polysaccharides ob-
tained by a sequential extraction of the same initial cell wall material.
The content of galacturonic acid is well predicted with a R2 of 0.94 from
the cell wall materials of two apple cultivars harvested at eight maturity
stages (Szymanska-Chargot et al., 2015). The degree of methylester-
ification of pectins is evaluated by FTIR with a R2 of 0.95 in the cell wall
pectic polysaccharide extracts of olive and pear pulp (Barros et al.,
2002) and from AIS of diverse fruit and vegetables (broccoli, carrot,
tomato, sugar beet, apple, mango) with R2 of 0.929 (Kyomugasho et al.,
2015). Concerning the prediction of total polyphenols by FTIR, good
results are observed with R2 of 0.865 on homogenates of blueberry
(Sinelli, Spinardi, Di Egidio, Mignani, & Casiraghi, 2008), R2 of 0.972
on centrifuged must and wine (Fragoso, Acena, Guasch, Mestres, &
Busto, 2011), R2 of 0.931 on freeze-dried powder of olive (Machado
et al., 2015), R2 of 0.65 on supernatant of raspberry (Andrianjaka-
Camps et al., 2015), R2 of 0.8281 on dry aqueous extracts of raw po-
tatoes (Ayvaz et al., 2016) and R2 of 0.96 on non-oxidized freeze-dried
apple powders (Bureau et al., 2012). However, all samples of this pa-
pers contained much higher concentrations of polyphenols than the
current AIS. The performance of model prediction is linked to the
content of compound of interest, its variability in the studied pheno-
typic variability and the sample preparation.
M.H.G. Canteri, et al. Carbohydrate Polymers 212 (2019) 186–196

Table 2 Talari et al., 2017) was positively correlated to phenolic com-

Prediction of cell wall components by ATR-FTIR directly from RM and AIS pounds, AUA and NCGlc;
(spectral area: 2000-800 cm−1). Means of 10 independent PLS modelling/va- (5) The region 1146-1147 cm−1, equivalent to v(C-O-C) glycosidic
lidation tests. bound between uronic acids (Alonso-Simón et al., 2011; Coimbra
Cell wall component Reference data Cross-validation data et al., 1998; Monti et al., 2013) was correlated to NCGlc and AIS
yield;
Mean SD LV R2 RPD (6) Positive values for phenolic compounds versus negative values for
Raw Materials (RM)
AIS yield at ˜1183 cm−1, relative to (C-O-C) from phenolic (Bich
Yield AIS DW* (mg g−1) 415.0 230.0 7-11 0.92 3.5 et al., 2009);
Total neutral sugars of AIS (mg 469.9 163.3 6-11 0.77 2.1 (7) Positive values for NCGlc and yield at near ˜1371 cm−1, associated
g−1) to C–H deformations of Lig (Abidi et al., 2014; Alonso-Simón et al.,
AUA (mg g−1) 190.8 82.8 11-13 0.75 1.9
2011; España et al., 2014; Monti et al., 2013);
MeOH (mg g−1) 16.8 12.1 9-13 0.80 2.2
DM (%) 43.5 20.5 8-14 0.79 2.1 (8) Positive values for degree of methylation corresponding to C]O
Starch (mg g−1) 120.5 241.7 6-11 0.83 2.3 stretching vibration from ionic carboxyl groups at 1616 cm−1
Phenolics (mg g−1) 0.5 0.6 8-13 0.73 2.0 (Kawabata, 1977);
Lig (mg g−1) 110.0 76.7 6-13 0.79 2.1
(9) Positive values for AUA and NCGlc versus negative values for AIS
AIS materials yield at 1705 cm−1 associated to C]O stretch of COOH of car-
Rha (mg g−1) 7.2 2.8 7-12 0.78 2.0 boxylic acid groups (España et al., 2014; Palacio et al., 2014) and
Fuc (mg g−1) 2.6 2.4 9-13 0.80 2.1
Ara (mg g−1) 48.1 32.8 9-11 0.91 3.2
(10) Positive values for DM, MeOH and total neutral sugars at
Xyl (mg g−1) 44.9 41.8 6-11 0.87 2.7 1740 cm−1, linked to C]O stretching vibration of alkyl ester from
Man (mg g−1) 14.4 9.8 8-12 0.80 2.2 esterified uronic acid (Alonso-Simón et al., 2011; Coimbra et al.,
Gal (mg g−1) 38.5 27.2 11-13 0.82 2.4 1998; Copikova, Cerna, Novotna, Kaasova, & Synytsya, 2001;
Total Glc (mg g−1) 314.4 194.1 5-9 0.95 4.4
España et al., 2014; Garside & Wyeth, 2003; Kawabata, 1977).
CGlc 192.2 79.9 5-12 0.57 1.6
NCGlc (mg g−1) 122.2 216.6 4-10 0.93 4.0
Total neutral sugars (mg g−1) 469.9 163.3 4-10 0.90 3.2 3.3. Reference methods: outlook and limitations
AUA (mg g−1) 190.8 82.8 3-11 0.86 2.6
MeOH (mg g−1) 16.8 12.1 4-9 0.94 4.1 As seen above, the cell wall characterization was complex and the
DM (%) 43.5 20.5 5-11 0.86 2.6
Starch (mg g−1) 120.5 241.7 5-7 0.96 4.9
quality of the data obtained by reference methods directly affected the
Phenolic (mg g−1) 0.5 0.6 4-14 0.78 2.1 prediction results obtained by ATR-FTIR spectroscopy. In this part,
Lig (mg g−1) 110.0 76.7 6-10 0.87 2.7 methods were reviewed to establish the link between the identified
difficulties to measure the cell wall compounds and the predicted re-
* On the total RM dry matter Abbreviations used AIS- alcohol insoluble so- sults.
lids, Ara: arabinose; AUA: anhydrous uronic acid; DM: degree of methylation;
The method chosen for the cell wall isolation was the hot alcohol
DW: dry weight; Fuc: fucose; Gal: galactose; Glc: glucose; LV: latent variables;
insoluble solids (Renard et al., 1990). The control of AIS material
Man: mannose; MeOH: methanol; NCGlc: non cellulosic glucose; Rha: rham-
nose; RMSECV: root-mean-square error of cross-validation; RPD: ratio of the
concerning its moisture content can be problematic due to the buildup
standard deviation; SD: standard deviation; Xyl: xylose. of static charge (Yeats, Vellosillo, Sorek, Ibáñez, & Bauer, 2016) and the
water in the samples promotes strong ATR-FTIR absorption bands in the
The main wavenumbers used in the models to assess the plant cell used spectral areas. To avoid this behavior in our sample, P2O5 atmo-
wall polysaccharides were in the region from 1750 cm−1 to 900 cm−1 sphere was used with its high water sorption capacity (Chen, Cao, &
as seen with the regression loadings (Fig. 5). This region corresponds to Mu, 2014). The AIS yield prediction from RM was good (Table 2) and
the relatively distinct signals for esterified and free uronic acids (1750 – for the other evaluated components directly in RM, R2 was around 0.8
1600 cm-1). and RPD lower than 2.3 due to low concentrations of individual AIS
Thirty different wavenumbers were associated to the different components in the raw powders. In addition, some components such as
models for the prediction of cell wall composition (Fig. 5) (Supple- phenolics can be abundant in the RM but only remain as traces or ar-
mentary Table 3). Among them, the following eight wavenumbers ap- tefacts in the AIS. This is notably the case of some phenolic compounds,
peared to be particularly significant and redundant in more than one which bind to AIS during its preparation or proteins, partially co-pre-
model: cipitated by the solvent used during AIS preparation (Renard, 2005;
Yapo, 2012).
(1) A signal at ˜921 cm−1, appeared to be correlated to the NCGlc, Concerning the neutral sugars from AIS samples, only three com-
starch and Lig. It corresponds to the skeletal mode vibrations of α- pounds among ten and their sum (Ara, Xyl, NCGlC and total neutral
1,4 glycosidic linkage (CeOeH) present in starch Fahey et al., sugars) presented good predictions with R2 above 0.87 and RPD above
2017; Kizil et al., 2002). The prediction of NCGlc and Lig is 2.5. In general, replicates presented a low standard deviation and so a
probably linked to their (negative) correlation with starch in the good reproducibility was obtained using the GC method to quantify
dataset (Fig. 2); alditol acetate derivatives. For the other sugars, only possible predic-
(2) The signal at ˜985 cm−1 corresponding to ρ(−CH-) in cellulose tion could be expected with RPD between 2 and 2.5 (Table 3). The
(Fahey et al., 2017; Garside & Wyeth, 2003) was negatively cor- results were not good because even though some spectral wavenumbers
related to neutral sugars and AIS yield; are specific of sugars, the reference method still depends on their hy-
(3) The region 1035–1041 cm−1 had positive values for Lig versus drolysis to the corresponding alditol (Sanz & Martínez-Castro, 2007).
negative values for phenolic compounds, MeOH, AUA and AIS Concerning cellulose, the hydrolysis of AIS should be adjusted ac-
yield; this range has been identified in arabinan, xyloglucan and cording to each material because the carbohydrate recovery can be
Lig (Alonso-Simón et al., 2011; Coimbra et al., 1998; Fahey et al., affected by their characteristics and physical structures as well as the
2017; Szymanska-Chargot & Zdunek, 2013; Toribio-Cuaya et al., time versus temperature of reaction and acid concentration (Selvendran
2014); & O’Neill, 1987). There is not a fixed pattern of hydrolysis performance
(4) The peak at 1075 cm−1, associated to v (C–O), v(C-C) from xy- (Zhou & Runge, 2014).
loglucan and phosphate (Szymanska-Chargot & Zdunek, 2013; Evaluating pectins in plant cell walls is a major challenge: they are
heterogeneous polymers showing different ratios of homogalacturonans

192
M.H.G. Canteri, et al.
Fig. 5. Regression loading distributions for the FT-IR wavenumbers of AIS yield from raw material spectra; total neutral sugars, non-cellulosic glucose, anhydrous
uronic acid, methanol and degree of methylation from AIS samples.
(smooth region) and rhamnogalacturonans backbone (hairy region). It
is impossible to extract them quantitatively from plant cell walls
without artefacts that modify their structure (Thibault, Saulnier,
Axelos, & Renard, 1991). This makes their correct analysis and there-
fore prediction by FT-IR particularly difficult. Pectins were quantified
here by their AUA content, their methanol content used to evaluate the
degree of methylation of the pectins and rhamnose content present in
the rhamnogalacturonans. The AUA prediction (Table 3) was good and
acceptable with the corresponding RPD close to the limit of 2.5
(R2 = 0.86 and RPD = 2.55). The determination of uronic acid content
is difficult due to the relatively high stability of the glycosyl uronic acid
193
Carbohydrate Polymers 212 (2019) 186–196
bonds and the instability of uronic acids under acidic conditions
(Kumar & Kumar, 2017), which means that colorimetric methods in
concentrated sulfuric acid must be used. These colorimetric assays have
many potential interference factors (Filisetti-Cozzi & Carpita, 1991):
H2SO4 concentration or rather contamination by water vapor, heating
duration and temperature, quality of mixing of the sample with the
viscous sulfuric acid, and interference by other components, such as red
discoloration from polyphenols and non-specific browning reaction
from neutral sugars. The difficulty of this analysis can be underlined by
the obtained high analytical relative standard deviation (pooled
SD = 11), in spite of strict control of conditions and experienced staff.
M.H.G. Canteri, et al.
Lack of precision of the reference methods precludes a good prediction
by the indirect ATR-FTIR method, even though spectra might contain
more precise information. In contrast, methanol, also from pectic
origin, presented an excellent prediction (R2 = 0.94 and RPD = 4.11,
see Table 3). The GCeMS method applied in this work (Renard &
Ginies, 2009) presents a high accuracy. Peaks are easily integrated due
to their good symmetry and intrinsic resolution. This method needs a
low amount of sample, and allows an easy and fast analysis of methanol
in comparison with other methods such as the one using HPLC
(Huisman, Oosterveld, & Schols, 2004). The DM prediction offered
good values of R2 (0.86) and of RPD (2.65). It was calculated with
values of MeOH content determined by GCeMS and of uronic acid
content determined by colorimetric analysis (Renard & Ginies, 2009).
Again, this highlights the importance of a reliable and high quality
reference analysis method, as the excellent results obtained for MeOH
are degraded by the lack of precision of AUA analysis.
Although the starch is not exactly a component of the cell wall, its
removal was not performed before cell wall characterization. Starch
prediction in AIS was excellent (R2 = 0.96 and RPD = 4.91), indicating
that the enzymatic method used for starch quantification is accurate in
spite of many manipulations. In this work, the iodine assay (Lugol’s
test) (Demiate & Jane, 2017) was only carried out to identify starch-
containing samples. The iodine assay was not retained for starch
quantification due to its various responses to amylose and amylopectin
(McCleary et al., 2006).
Phenolic compounds were only present in very small concentrations
in AIS (generally below than 1 mg g−1 AIS). The ATR-FTIR spectro-
scopy was then limited to predict their content (R2 = 0.78 and
RPD = 2.09). During the AIS fraction preparation, phenolics were re-
moved but some of them could bond to the cell wall and were con-
sidered as an artefact.
The lignin prediction by ATR-FTIR was considered to be good and
acceptable (R2 = 0.87 and RPD = 2.73). However, the determination of
the native lignin is not accurate due to the difficulties met to degrade
this compound before analysis. Standardized procedures emphasize the
complete dissolution of the cell wall components, without formation of
non-lignin UV absorbing-compounds (Macheix, Fleuriet, & Sarni-
Manchado, 2006). The oxidative degradation of structural poly-
saccharides during cell wall incubation with acid may contribute to
promote overestimation of the lignin content. Another limitation is to
obtain a well-defined lignin standard for calibration (Moreira-Vilar
et al., 2014). This would not affect the PLS results but leads to a sys-
tematic error (defect in accuracy).
4. Conclusion
Coupled with PCA, ATR-FTIR allowed a good discrimination be-
tween samples rich in starch from samples with substantial amounts of
lignin and hemicelluloses or methyl esterified pectins. Moreover, by
coupling ATR-FTIR with PLS regressions, the yield of AIS can be di-
rectly predicted from the freeze-dried raw materials. It makes possible
to determine indirectly the total dietary fiber in human nutrition and
animal feed using a very simple method. This indicates that such ATR-
FTIR methods might be developed to extend fiber analysis to control the
product quality, after a calibration with an AOAC method.
ATR-FTIR performed on AIS materials also allowed the prediction of
the content of total glucose, NCGlc, Xyl, AUA, MeOH, DM, arabinose,
Lig and starch (RPD > 2.5). ATR-FTIR could therefore be used to es-
timate cell wall components such as cellulose, pectins (with a good
accuracy for degree of methylation and indications of side-chains
(arabinose)), and Lig. By estimating Xyl, some indications on hemi-
celluloses might be deduced. From this, it is possible indirectly to infer
the composition of total dietary fibers, facilitating its technological
application, through the correlation of its macromolecular properties
with its physicochemical and mechanical properties. Also, it could in-
fluence its utilization as a prebiotic or promoter of intestinal flow.
194
Carbohydrate Polymers 212 (2019) 186–196
The time of analysis was substantially shortened and simplified
(samples directly put on an ATR crystal and 30 s are needed for each
spectrum acquisition) compared to the conventional biochemical
methods for each component. The proposed models were generic or
multi-species insofar as 29 species of fruits and vegetables were in-
vestigated here to cover a large range of variability for each studied
compound of cell wall. These results can be successfully implemented
on ATR-FTIR spectra from raw and AIS materials, however the authors
suggest the starch removal before spectrum acquisition on AIS samples
to avoid interferences.
Acknowledgements
The authors thank Line Touloumet, Marielle Boge and Romain Bott
(INRA, Avignon) for their technical help, Flavia Henrique for the pre-
paration of yerba mate samples and their respective Institutions, which
provided necessary resources to perform this study. Maria H. G. Canteri
was supported by CAPES- Coordenação de Aperfeiçoamento de Pessoal
de Nível Superior (99999.006882/2014-05) from Ministry of Education
(Brazil) and CNPq - Conselho Nacional de Desenvolvimento Científico e
Tecnológico (487621/2012-3–APQ) from Ministry of Science and
Technology (Brazil).
Appendix A. Supplementary data
Supplementary material related to this article can be found, in the
online version, at doi:https://doi.org/10.1016/j.carbpol.2019.02.021.
References
Abidi, N., Cabrales, L., & Haigler, C. H. (2014). Changes in the cell wall and cellulose
content of developing cotton fibers investigated by FTIR spectroscopy. Carbohydrate
Polymers, 100, 9–16.
Alonso-Simón, A., García-Angulo, P., Mélida, H., Encina, A., Álvarez, J. M., & Acebes, J. L.
(2011). The use of FTIR spectroscopy to monitor modifications in plant cell wall
architecture caused by cellulose biosynthesis inhibitors. Plant Signaling & Behavior,
6(8), 1104–1110.
Amarasekara, A. S. (2013). Enzymatic Hydrolysis of Cellulose and Hemicellulose. In A. S.
Amarasekara (Ed.). Handbook of Cellulosic Ethanol (pp. 219–246). USA: John Wiley .
Andrianjaka-Camps, Z. N., Baumgartner, D., Camps, C., Guyer, E., Arrigoni, E., & Carlen,
C. (2015). Prediction of raspberries puree quality traits by Fourier transform infrared
spectroscopy. LWT-Food Science and Technology, 63(2), 1056–1062.
Ayvaz, H., Bozdogan, A., Giusti, M. M., Mortas, M., Gomez, R., & Rodriguez-Saona, L. E.
(2016). Improving the screening of potato breeding lines for specific nutritional traits
using portable mid-infrared spectroscopy and multivariate analysis. Food Chemistry,
211, 374–382.
Barros, A. S., Mafra, I., Ferreira, D., Cardoso, S., Reis, A., da Silva, J. L., ... Coimbra, M. A.
(2002). Determination of the degree of methylesterification of pectic polysaccharides
by FT-IR using an outer product PLS1 regression. Carbohydrate Polymers, 50(1),
85–94.
Bich, V. T., Thuy, N. T., Binh, N. T., Huong, N. T. M., Yen, P. N. D., & Luong, T. T. (2009).
Structural and spectral properties of curcumin and metal-curcumin complex derived from
turmeric (Curcuma longa). Physics and engineering of new materials. Berlin, Heidelberg:
Springer271–278.
Bichara, L. C., Alvarez, P. E., Bimbi, M. V. F., Vaca, H., Gervasi, C., & Brandán, S. A.
(2016). Structural and spectroscopic study of a pectin isolated from citrus peel by
using FTIR and FT-Raman spectra and DFT calculations. Infrared Physics &
Technology, 76, 315–327.
Bociek, S. M., & Welti, D. (1975). The quantitative analysis of uronic acid polymers by
infrared spectroscopy. Carbohydrate Research, 42(2), 217–226.
Box, G. E., Hunter, W. G., & Hunter, J. S. (1978). Statistics for experimenters, an introduction
to design, data analysis and model building. USA: John Wiley & Sons352.
Brahem, M., Renard, C. M., Gouble, B., Bureau, S., & Le Bourvellec, C. (2017).
Characterization of tissue specific differences in cell wall polysaccharides of ripe and
overripe pear fruit. Carbohydrate Polymers, 156, 152–164.
Bureau, S., Ścibisz, I., Le Bourvellec, C., & Renard, C. M. (2012). Effect of sample pre-
paration on the measurement of sugars, organic acids, and polyphenols in apple fruit
by mid-infrared spectroscopy. Journal of Agricultural and Food Chemistry, 60(14),
3551–3563.
Caffall, K. H., & Mohnen, D. (2009). The structure, function, and biosynthesis of plant cell
wall pectic polysaccharides. Carbohydrate Research, 344(14), 1879–1900.
Carpita, N. C., & Gibeaut, D. M. (1993). Structural models of primary cell walls in
flowering plants: Consistency of molecular structure with the physical properties of
the walls during growth. The Plant Journal, 3(1), 1–30.
Chatjigakis, A. K., Pappas, C., Proxenia, N., Kalantzi, O., Rodis, P., & Polissiou, M. (1998).
FT-IR spectroscopic determination of the degree of esterification of cell wall pectins
M.H.G. Canteri, et al.
from stored peaches and correlation to textural changes. Carbohydrate Polymers,
37(4), 395–408.
Chen, Y., Cao, Y., & Mu, T. (2014). A new application of acetate‐based ionic liquids:
Potential usage as drying materials. Chemical Engineering & Technology, 37(3),
527–534.
Chylińska, M., Szymańska-Chargot, M., & Zdunek, A. (2016). FT-IR and FT-Raman
characterization of non-cellulosic polysaccharides fractions isolated from plant cell
wall. Carbohydrate Polymers, 154, 48–54.
Coates, J. (2000). Interpretation of infrared spectra, a practical approach. Encyclopedia of
analytical chemistry, 12, 10815–10837.
Coimbra, M. A., Barros, A., Barros, M., Rutledge, D. N., & Delgadillo, I. (1998).
Multivariate analysis of uronic acid and neutral sugars in whole pectic samples by FT-
IR spectroscopy. Carbohydrate Polymers, 37(3), 241–248.
Copikova, J., Cerna, M., Novotna, M., Kaasova, J. I. T. K. A., & Synytsya, A. (2001).
Application of FT-IR spectroscopy in detection of food hydrocolloids confectionery
jellies and in food supplements. Czech Journal of Food Sciences, 19(2), 51–56.
Costa, G., & Plazanet, I. (2016). Plant cell wall, a challenge for its characterisation.
Advances in Biological Chemistry, 6(03), 70.
Demiate, I. M., & Jane, J.-L. (2017). Obtenção e Análise de Propriedades Funcionais de
Amidos: Fundamentos e Aplicações. In D. Granato, & D. Nunes (Eds.). Análises
Químicas, Propriedades Funcionais e Controle de Qualidade de Alimentos e Bebidas: Uma
Abordagem Teórico-Prática (pp. 27–62). Brasil: Elsevier.
Demiate, I. M., Dupuy, N., Huvenne, J. P., Cereda, M. P., & Wosiacki, G. (2000).
Relationship between baking behavior of modified cassava starches and starch che-
mical structure determined by FTIR spectroscopy. Carbohydrate Polymers, 42(2),
149–158.
España, L., Heredia‐Guerrero, J. A., Segado, P., Benítez, J. J., Heredia, A., & Domínguez,
E. (2014). Biomechanical properties of the tomato (Solanum lycopersicum) fruit cuticle
during development are modulated by changes in the relative amounts of its com-
ponents. The New Phytologist, 202(3), 790–802.
Fahey, L. M., Nieuwoudt, M. K., & Harris, P. J. (2017). Predicting the cell-wall compo-
sitions of Pinus radiata (radiata pine) wood using ATR and transmission FTIR spec-
troscopies. Cellulose, 24(12), 5275–5293.
Faix, O. (1991). Classification of lignins from different botanical origins by FT-IR spec-
troscopy. Holzforschung-International Journal of the Biology, Chemistry, Physics and
Technology of Wood, 45(s1), 21–28.
Fellah, A., Anjukandi, P., Waterland, M. R., & Williams, M. A. (2009). Determining the
degree of methylesterification of pectin by ATR/FT-IR: Methodology optimisation
and comparison with theoretical calculations. Carbohydrate Polymers, 78(4),
847–853.
Ferreira, D., Barros, A., Coimbra, M. A., & Delgadillo, I. (2001). Use of FT-IR spectroscopy
to follow the effect of processing in cell wall polysaccharide extracts of a sun-dried
pear. Carbohydrate Polymers, 45(2), 175–182.
Filippov, M. P. (1992). Practical infrared spectroscopy of pectic substances. Food
Hydrocolloids, 6(1), 115–142.
Filippov, M. P., & Chernei, G. V. (2002). Determination of the content and degree of
esterification of uronic acids in plant tissues and products of their processing. Applied
Biochemistry and Microbiology, 38(1), 68–71.
Filippov, M. P., & Kohn, R. (1975). Determination of esterification degree of carboxyl
groups of pectin with methanol by means of infra-red spectroscopy. Chemicke Zvesti,
29, 88–91.
Filisetti-Cozzi, T. M., & Carpita, N. C. (1991). Measurement of uronic acids without in-
terference from neutral sugars. Analytical Biochemistry, 197(1), 157–162.
Fragoso, S., Acena, L., Guasch, J., Mestres, M., & Busto, O. (2011). Quantification of
phenolic compounds during red winemaking using FT-MIR spectroscopy and PLS-
regression. Journal of Agricultural and Food Chemistry, 59(20), 10795–10802.
Garside, P., & Wyeth, P. (2003). Identification of cellulosic fibres by FTIR spectroscopy-
thread and single fibre analysis by attenuated total reflectance. Studies in
Conservation, 48(4), 269–275.
Hayashi, T., & Kaida, R. (2011). Functions of xyloglucan in plant cells. Molecular Plant,
4(1), 17–24.
Houben, K., Jolie, R. P., Fraeye, I., Van Loey, A. M., & Hendrickx, M. E. (2011).
Comparative study of the cell wall composition of broccoli, carrot, and tomato:
Structural characterization of the extractable pectins and hemicelluloses.
Carbohydrate Research, 346(9), 1105–1111.
Huck, C. W. (2015). Advances of infrared spectroscopy in natural product research.
Phytochemistry Letters, 11, 384–393.
Huisman, M. M. H., Oosterveld, A., & Schols, H. A. (2004). Fast determination of the
degree of methyl esterification of pectins by head-space GC. Food Hydrocolloids,
18(4), 665–668.
Jones, J. M. (2014). CODEX-aligned dietary fiber definitions help to bridge the ‘fiber gap’.
Nutrition Journal, 13(1), 34.
Kačuráková, M., & Wilson, R. H. (2001). Developments in mid-infrared FT-IR spectro-
scopy of selected carbohydrates. Carbohydrate Polymers, 44(4), 291–303.
Kawabata, A. (1977). Studies on chemical and physical properties of pectic substances
from fruits. Memoirs of the Tokyo University of Agriculture..
Kizil, R., Irudayaraj, J., & Seetharaman, K. (2002). Characterization of irradiated starches
by using FT-Raman and FTIR spectroscopy. Journal of Agricultural and Food Chemistry,
50(14), 3912–3918.
Kosmala, M., Milala, J., Kołodziejczyk, K., Markowski, J., Zbrzeźniak, M., & Renard, C. M.
(2013). Dietary fiber and cell wall polysaccharides from plum (Prunus domestica L.)
fruit, juice and pomace: Comparison of composition and functional properties for
three plum varieties. Food Research International, 54(2), 1787–1794.
Kumar, P., & Kumar, V. (2017). Estimation of uronic acids using diverse approaches and
monosaccharide composition of alkali soluble polysaccharide from Vitex negundo
Linn. Carbohydrate Polymers, 165, 205–212.
195
Carbohydrate Polymers 212 (2019) 186–196
Kurz, C., Leitenberger, M., Carle, R., & Schieber, A. (2010). Evaluation of fruit authen-
ticity and determination of the fruit content of fruit products using FT-NIR spectro-
scopy of cell wall components. Food Chemistry, 119(2), 806–812.
Kyomugasho, C., Christiaens, S., Shpigelman, A., Van Loey, A. M., & Hendrickx, M. E.
(2015). FT-IR spectroscopy, a reliable method for routine analysis of the degree of
methylesterification of pectin in different fruit-and vegetable-based matrices. Food
Chemistry, 176, 82–90.
Lee, L. C., Liong, C. Y., & Jemain, A. A. (2017). A contemporary review on Data pre-
processing (DP) practice strategy in ATR-FTIR spectrum. Chemometrics and Intelligent
Laboratory Systems, 163, 64–75.
Liu, Q., Luo, L., & Zheng, L. (2018). Lignins: Biosynthesis and biological functions in
plants. International Journal of Molecular Sciences, 19(2), 335.
López-Sánchez, M., Ayora-Cañada, M. J., & Molina-Díaz, A. (2009). Olive fruit growth
and ripening as seen by vibrational spectroscopy. Journal of Agricultural and Food
Chemistry, 58(1), 82–87.
Macheix, J. J., Fleuriet, A., & Sarni-Manchado, P. (2006). In P. Coordonnatrices Sarni-
Manchado, V. Cheynier, & TEC et DOC (Eds.). Composés phénoliques dans la plante-
structure, biosynthèse, répartition et rôles. Les polyphénols en agroalimentaire (pp. 390–
399). Paris: Lavoisier.
McCann, M. C., Defernez, M., Urbanowicz, B. R., Tewari, J. C., Langewisch, T., Olek, A., ...
Carpita, N. C. (2007). Neural network analyses of infrared spectra for classifying cell
wall architectures. Plant Physiology, 143(3), 1314–1326.
McCleary, B. V., Charnock, S. J., Rossiter, P. C., O’Shea, M. F., Power, A. M., & Lloyd, R.
M. (2006). Measurement of carbohydrates in grain, feed and food. Journal of the
Science of Food and Agriculture, 86(11), 1648–1661.
Monti, F., Dell’Anna, R., Sanson, A., Fasoli, M., Pezzotti, M., & Zenoni, S. (2013). A
multivariate statistical analysis approach to highlight molecular processes in plant
cell walls through ATR FT-IR microspectroscopy: The role of the α-expansin
PhEXPA1 in Petunia hybrida. Vibrational Spectroscopy, 65, 36–43.
Moreira-Vilar, F. C., de Cássia Siqueira-Soares, R., Finger-Teixeira, A., de Oliveira, D. M.,
Ferro, A. P., da Rocha, G. J., ... Ferrarese-Filho, O. (2014). The acetyl bromide
method is faster, simpler and presents best recovery of lignin in different herbaceous
tissues than klason and thioglycolic acid methods. PloS One, 9(10), e110000.
Nicolai, B. M., Beullens, K., Bobelyn, E., Peirs, A., Saeys, W., Theron, K. I., ... Lammertyn,
J. (2007). Nondestructive measurement of fruit and vegetable quality by means of
NIR spectroscopy: A review. Postharvest Biology and Technology, 46(2), 99–118.
Niklas, K. J. (2004). The cell walls that bind the tree of life. AIBS Bulletin, 54(9), 831–841.
Nunes, C. A., Alvarenga, V. O., de Souza Sant’Ana, A., Santos, J. S., & Granato, D. (2015).
The use of statistical software in food science and technology: Advantages, limita-
tions and misuses. Food Research International, 75, 270–280.
Palacio, S., Aitkenhead, M., Escudero, A., Montserrat-Martí, G., Maestro, M., & Robertson,
A. J. (2014). Gypsophile chemistry unveiled: Fourier transform infrared (FTIR)
spectroscopy provides new insight into plant adaptations to gypsum soils. PloS One,
9(9), e107285.
Popper, Z. A., Michel, G., Hervé, C., Domozych, D. S., Willats, W. G., Tuohy, M. G., ...
Stengel, D. B. (2011). Evolution and diversity of plant cell walls: From algae to
flowering plants. Annual Review of Plant Biology, 62, 567–590.
Renard, C. M. (2005). Variability in cell wall preparations: Quantification and comparison
of common methods. Carbohydrate Polymers, 60(4), 515–522.
Renard, C. M., & Ginies, C. (2009). Comparison of the cell wall composition for flesh and
skin from five different plums. Food Chemistry, 114(3), 1042–1049.
Renard, C. M. G. C., Voragen, A. G. J., Thibault, J. F., & Pilnik, W. (1990). Studies on
apple protopectin: I. Extraction of insoluble pectin by chemical means. Carbohydrate
Polymers, 12(1), 9–25.
Robic, A., Bertrand, D., Sassi, J. F., Lerat, Y., & Lahaye, M. (2009). Determination of the
chemical composition of ulvan, a cell wall polysaccharide from Ulva spp.(Ulvales,
Chlorophyta) by FT-IR and chemometrics. Journal of Applied Phycology, 21(4),
451–456.
Ruiz-Matute, A. I., Hernandez-Hernandez, O., Rodriguez-Sanchez, S., Sanz, M. L., &
Martinez-Castro, I. (2011). Derivatization of carbohydrates for GC and GC–MS ana-
lyses. Journal of Chromatography B, 879(17-18), 1226–1240.
Sanz, M. L., & Martínez-Castro, I. (2007). Recent developments in sample preparation for
chromatographic analysis of carbohydrates. Journal of Chromatography A, 1153(1-2),
74–89.
Schols, H. A., Bakx, E. J., Schipper, D., & Voragen, A. G. (1995). A xylogalacturonan
subunit present in the modified hairy regions of apple pectin. Carbohydrate Research,
279, 265–279.
Ścibisz, I., Reich, M., Bureau, S., Gouble, B., Causse, M., Bertrand, D., ... Renard, C. M.
(2011). Mid-infrared spectroscopy as a tool for rapid determination of internal
quality parameters in tomato. Food Chemistry, 125(4), 1390–1397.
Selvendran, R. R., & O’Neill, M. A. (1987). Isolation and analysis of cell walls from plant
material. Methods of Biochemical Analysis, 25–153.
Sevenou, O., Hill, S. E., Farhat, I. A., & Mitchell, J. R. (2002). Organisation of the external
region of the starch granule as determined by infrared spectroscopy. International
Journal of Biological Macromolecules, 31(1–3), 79–85.
Sinelli, N., Spinardi, A., Di Egidio, V., Mignani, I., & Casiraghi, E. (2008). Evaluation of
quality and nutraceutical content of blueberries (Vaccinium corymbosum L.) by near
and mid-infrared spectroscopy. Postharvest Biology and Technology, 50(1), 31–36.
Singleton, V. L., & Rossi, J. A. (1965). Colorimetry of total phenolics with phosphomo-
lybdic-phosphotungstic acid reagents. American Journal of Enology and Viticulture,
16(3), 144–158.
Sørensen, I., Domozych, D., & Willats, W. G. (2010). How have plant cell walls evolved?
Plant Physiology, 153(2), 366–372.
Szymanska-Chargot, M., Chylinska, M., Kruk, B., & Zdunek, A. (2015). Combining FT-IR
spectroscopy and multivariate analysis for qualitative and quantitative analysis of the
cell wall composition changes during apples development. Carbohydrate Polymers,
M.H.G. Canteri, et al.
115, 93–103.
Szymańska-Chargot, M., Chylińska, M., Gdula, K., Kozioł, A., & Zdunek, A. (2017).
Isolation and characterization of cellulose from different fruit and vegetable pomaces.
Polymers, 9(10), 495.
Szymanska-Chargot, M., & Zdunek, A. (2013). Use of FT-IR spectra and PCA to the bulk
characterization of cell wall residues of fruits and vegetables along a fraction process.
Food Biophysics, 8(1), 29–42.
Talari, A. C. S., Martinez, M. A. G., Movasaghi, Z., Rehman, S., & Rehman, I. U. (2017).
Advances in Fourier transform infrared (FTIR) spectroscopy of biological tissues.
Applied Spectroscopy Reviews, 52(5), 456–506.
Tamaki, Y., & Mazza, G. (2011). Rapid determination of carbohydrates, ash, and ex-
tractives contents of straw using attenuated total reflectance Fourier transform mid-
infrared spectroscopy. Journal of Agricultural and Food Chemistry, 59(12), 6346–6352.
Thibault, J. F., Saulnier, L., Axelos, M. A., & Renard, C. M. (1991). Difficultés
expérimentales de l’étude des macromolecules pectiques. Bulletin de la Société
Botanique de France. Actualités Botaniques, 138(3-4), 319–337.
Toribio-Cuaya, H., Pedraza-Segura, L., Macías-Bravo, S., Gonzalez-García, I., Vasquez-
Medrano, R., & Favela-Torres, E. (2014). Characterization of lignocellulosic biomass
using five simple steps. Journal of Chemical, Biological and Physical Sciences, 4(5),
28–47.
Türker-Kaya, S., & Huck, C. W. (2017). A review of mid-infrared and near-infrared
196
Carbohydrate Polymers 212 (2019) 186–196
imaging: Principles, concepts and applications in plant tissue analysis. Molecules,
22(1), 168.
Udén, P. (2009). Estimating residual starch from in vitro fermentations by Fourier
transform mid-IR transmission spectroscopy. Animal Feed Science and Technology,
152(1-2), 133–140.
Voragen, A. G., Coenen, G. J., Verhoef, R. P., & Schols, H. A. (2009). Pectin, a versatile
polysaccharide present in plant cell walls. Structural Chemistry, 20(2), 263.
Warren, F. J., Gidley, M. J., & Flanagan, B. M. (2016). Infrared spectroscopy as a tool to
characterise starch ordered structure—a joint FTIR–ATR, NMR, XRD and DSC study.
Carbohydrate Polymers, 139, 35–42.
Xu, F., Yu, J., Tesso, T., Dowell, F., & Wang, D. (2013). Qualitative and quantitative
analysis of lignocellulosic biomass using infrared techniques: A mini-review. Applied
Energy, 104, 801–809.
Yapo, B. M. (2012). On the colorimetric-sulfuric acid analysis of uronic acids in food
materials: Potential sources of discrepancies in data and how to circumvent them.
Food Analytical Methods, 5(2), 195–215.
Yeats, T., Vellosillo, T., Sorek, N., Ibáñez, A. B., & Bauer, S. (2016). Rapid determination
of cellulose, neutral sugars, and uronic acids from plant cell walls by one-step two-
step hydrolysis and HPAEC-PAD. Plant Physiology.
Zhou, S., & Runge, T. M. (2014). Validation of lignocellulosic biomass carbohydrates
determination via acid hydrolysis. Carbohydrate Polymers, 112, 179–185.
 Update
Carbohydrate Polymers
Volume 235, Issue , 1 May 2020, Page

DOI: https://doi.org/10.1016/j.carbpol.2020.115960
 Carbohydrate Polymers 235 (2020) 115960

Contents lists available at ScienceDirect

Carbohydrate Polymers
journal homepage: www.elsevier.com/locate/carbpol

Corrigendum

Corrigendum to “ATR-FTIR spectroscopy to determine cell wall T

composition: Application on a large diversity of fruits and vegetables”
[Carbohydr. Polym. 212 (May) (2019), 186–196]
Maria H.G. Canteria,*, Catherine M.G.C Renardb, Carine Le Bourvellecb, Sylvie Bureaub
a
Department of Chemistry and Biology-DAQBI, Universidade Tecnológica Federal do Paraná, UTFPR, Brazil
b
UMR408 SQPOV, Sécurité et Qualité des Produits d’Origine Végétale, INRA, Avignon University, F-84000 Avignon, France

The authors regret to inform that the published version of the
Supplementary Table 1 presented some errors not detected in the
proofs. These differences were found when the authors wished to reuse
the replicates of the original data.
The authors have checked all values and corrected them, when
necessary, according to the averages of the replicates for each sample,
also used for the multivariate analysis. In addition, some missing
average values were inserted in the YFW column, as well as an extra
explanation in the Table caption for calculating the NCGlc values for
samples containing starch.
Please find enclosed, in the next page, the Supplementary Table 1
after corrections.
The authors would like to apologise for any inconvenience caused.
Supplementary Table 1 Composition of AIS extracted from fruit and
vegetables (mg g−1 dry weight, except for DM express in %).

DOI of original article: https://doi.org/10.1016/j.carbpol.2019.02.021

Corresponding author at: Federal University of Technology, Department of Chemistry and Biology-DAQBI, Francisco Beltrão, Linha Santa Bárbara, s/n, Paraná,
⁎

Brazil.
E-mail addresses: canteri@utfpr.edu.br, canteri.mhg@gmail.com (M.H.G. Canteri).

https://doi.org/10.1016/j.carbpol.2020.115960

Available online 18 February 2020

0144-8617/ © 2020 Elsevier Ltd. All rights reserved.
M.H.G. Canteri, et al. Carbohydrate Polymers 235 (2020) 115960

Tissues Codesa YDW YFW Rha Fuc Ara Xyl Man Gal CGlc NCGlc AUA MeOH DM Stch PC Lig

Flesh APPf (f) 149 23 9 7 81 38 14 55 212 151 221 28 68 102 0.15 40

BANf(f/d) 584 162 3 0 6 5 10 2 137 0b 68 4 31 772 0.04 60
PITA (d) 245 55 7 5 105 34 20 58 164 19 250 20 44 0 0.45 99
BROC (f) 424 59 10 4 103 30 18 59 209 10 233 16 39 0 0.05 74
CARO (f) 246 26 12 0 32 12 17 58 187 17 238 34 78 0 0.06 24
CAUL (f) 388 41 7 3 74 20 11 39 201 9 151 14 52 0 0.04 94
EGGP (d) 659 41 13 0 48 57 29 123 322 14 210 11 29 0 0.15 103
GUAB(d) 490 94 6 3 52 92 13 40 275 5 136 9 37 0 0.53 247
GUAV(d) 609 91 6 3 39 153 0 17 255 6 128 7 32 0 0.36 256
JABU (d) 500 82 7 1 45 17 13 45 258 0b 126 8 36 713 1.81 165
CITR (f) 169 18 13 4 54 28 19 44 143 12 311 41 72 0 1.25 98
ORAN(f) 132 18 13 3 96 20 17 81 134 15 328 29 49 0 2.19 48
PAPA (d) 354 40 8 3 8 30 20 19 216 10 205 10 27 0 0.10 197
PEAf (f) 101 15 9 7 110 35 19 68 207 11 284 34 66 0 0.04 63
KAKI (d) 319 65 7 3 33 119 10 33 146 10 193 16 46 0 0.67 177
ANAN (f) 76 11 6 5 73 134 22 52 263 14 105 3 14 0 0.09 118
PAMP (f) 140 15 8 4 46 30 19 41 172 14 356 38 58 0 0.05 48
ARAC (d) 623 80 5 0 28 146 0 8 209 11 120 5 25 0 0.69 268
BBER (d) 537 54 5 0 13 111 0 9 239 8 129 3 13 0 0.64 194
PUMP (f) 253 12 8 0 13 18 14 7 225 21 280 29 56 0 0.04 71
TOMA (f) 240 23 6 0 25 30 35 35 220 11 231 25 60 0 0.33 38
Leaf CELE (f) 353 38 8 0 73 14 13 27 162 6 262 35 74 0 0.06 82
LEEK (f) 298 42 6 2 27 23 13 113 180 7 187 21 61 0 0.04 108
SALA (f) 431 24 10 2 25 24 15 34 129 8 277 16 32 0 0.07 83
SPIN (f) 472 66 7 0 33 10 3 17 93 0 144 12 44 0 0.04 81
HERB (d) 554 211 6 2 51 50 9 29 113 21 114 7 30 22 0.52 218
Peel APPp (f) 344 72 7 4 83 19 13 45 124 203 196 19 55 164 0.2 44
BANp (d) 764 94 0 0 42 35 22 19 124 27 119 8 37 18 1.48 175
PEAp (f) 235 41 8 4 95 33 16 53 158 9 297 31 58 2 0.10 58
PASS (d) 644 138 9 5 10 33 35 27 313 24 281 37 83 4 0.04 53
Tuber POTA (f) 805 162 5 0 7 2 0 34 48 0b 49 1 13 845 0.10 66
PATA (d) 698 213 5 0 7 4 0 9 135 0b 63 2 13 778 0.09 46
Total meanc 435 43 7 2 48 45 14 39 183 129 183 16 44 156 0.40 113
Pooled SD n.d. n.d. 1 0.5 5 3 1 5 42 17 9 0.4 2 23 0.12 14
Tissues F-value 4.8 8.5 4.2 5.8 2.5 2.2 2.7 1.3 22.5 10.1 3.9 5.3 3.8 10.8
*** *** *** *** * * * Ns *** *** *** *** *** ***
Specie F-value 6.4 12.0 16.3 48.8 11.2 22.2 7.1 344.2 62.1 71.1 19.6 127.9 13.2 37.48
*** *** *** *** *** *** *** *** *** *** *** *** *** ***

Abbreviations used: acodes for species described in Table 1, Ara—arabinose, AUA—anhydrouronic acid, CGlc—cellulosic glucose (calculated by glucose
from Saeman hydrolysis – glucose from simple hydrolysis), DM—degree of methylation, Fuc—fucose, Gal—galactose, Lig—lignin, Man—mannose,
MeOH—methanol, NCGlc—non cellulosic glucose: for samples without starch=(glucose from simple hydrolysis) and bfor starchy samples=(simple hydrolysis
– starch); n.d.—not determined, PC— phenolic compounds, expressed as gallic acid express; Rha—rhamnose, Stch—starch, Xyl—xylose, YDW—yield dry
weight; YFW—yield fresh weight; (d)-dried powder before freeze drying, (f)-fresh before freeze drying; *, **, ***: nonsignificant or significant at p < 0.05,
0.01, 0.001, respectively; ctotal mean of all replicates.