Chitosan is a natural biodegradable cationic polymer with a promising potency for different applications. Its use is limited by standardization problems related to chitosan nature. The aim of this work was to compare different methods of molecular weight (MW) and deacetylation degree (DD) determination using two chitosan series varied in MW and DD. The absolute DD appeared to be accurately determined by NMR, first derivative of UV-absorbance spectra analysis, or conductometric titration. Contrary to DD, MW estimated by viscosity, size exclusion chromatography, or asymmetric field flow fractionation did not provide comparable absolute MW values. At the same time, correlation between all the methods was extremely high (Pearson’s coefficient of correlation, r~1) showing that each of them is sensitive to the change in chitosan MW. Taken collectively it can be concluded that chitosan behaves in different conditions as a molecule with different MW. As MW determined by any of these methods and viscosity correlate highly significantly, viscosity as a directly determined value is recommended to be used as the main chitosan characteristic.
Deacetylated chitin derivative chitosan is an attractive biopolymer due to its biocompatibility, biodegradability, low toxicity, and abundant reactive groups. Nevertheless practical application of chitosan and its derivatives is limited by standardization problems. These difficulties are caused by the nature of chitosan (variety of sources, dispersity, impurities) as well as the variations in molecular weight (MW) and acetylation degree (DD) determined by different methods. In 2011 German company Heppe Medical Chitosan GmbH sent chitosan samples to eleven European laboratories in order to independently determinate MW and DD. The data on DD were consistent while MW of the same sample varied in a broad range  resulted in the conclusion about intrinsic problems of chitosan standardization.
DD can be determined by three different types of methods: spectral (1Н-NMR, 13С-NMR, 15N-NMR, UV, IR); analytical (conductometric and potentiometric titration, ninhydrin method); and degradation (elemental analysis, complete hydrolysis of chitosan followed by colorimetric or HPLC analysis, differential thermal analysis) . There are no published papers which compared all these methods, however a few of them were compared by many groups [3-5].
Determination of chitosan MW, in our opinion, is a more complicated problem. This relates to the dispersity of the polymer and its tendency to self-association in solutions [6,7]. A variety of methods currently used to estimate chitosan MW (size exclusion chromatography (SEC), SEC coupled with MALS detector, viscometry, osmometry, dynamic and static light scattering, end group analyses, analytical ultracentifugation, et cetera) produce results which are often poorly compatible [8,9].
The problem of dispersity can be partially solved by fractionation methods, such as ones developed by Giddings [10,11] and nowadays realized by Post nova Analytics company (Germany). This group of methods is based on the separation of macromolecules as well as nanoparticles in a flow of liquid under the influence of perpendicular acting forces of different nature (sedimentation or thermal field, perpendicular flow etc.). In this paper a comparative study and correlation analysis of DD and MW, obtained by different methods including asymmetric field flow fractionation, were carried out.
aMethods used to determine Mark-Houwink coefficients in
b-dk and α–empirical Mark-Houwink coefficients according to
eDD was calculated using equation 1.
Protocol 1 was proposed for chitosan with different DD; protocols
2 and 3 describe the analysis of chitosan with high DD.
Chitosan (А1, 200 kDa, DD 90%) (Table 1) and chitosan
1000 kDa, DD 87% were purchased from ZAO «Bioprogress»,
Russia. Chitosan 1000 kDa was used to obtain samples B and C
(Table 1) by acidic hydrolysis. Glacial acetic acid (AcOH), sodium
hydroxide (NaOH), hydrochloric acid (HCl), acetic anhydride
(Ac2O), ammonium hydroxide (NH4OH) (reagent-grade),
methanol (MeOH) (Sigma-Aldrich, USA) were used as received.
Chitosan was purified as described earlier . To obtain
chitosan samples with different DD and the same MW, A1 was
re-acetylated as described earlier . Chitosan with MW ~50
kDa (sample B) and ~20 kDa (sample C) were prepared by acidic
hydrolysis of chitosan 1000 kDa as described .
Conductometric titration: Analytical sample (0.1g) of
chitosan was dried to a constant weight in microwave oven.
Standard 0.1MHCl solution was prepared using standard
titrimetric substance. Concentration of NaOH solution was
evaluated with standard 0.1M HCl solution. Sample was
dissolved in 10 ml 0.1М HCl and then diluted with 40 ml of
distilled H2O. For titration curve plotting every 30 sec 100 μl
of NaOH titrated solution were added.
Proton nuclear magnetic resonance (1Н-NMR): Proton
spectra of chitosan and its derivatives were registered by
Avance III (Bruker, Germany) spectrometer (operating
frequency for the protons-600 MHz, temperature 30 ˚C). The
samples were prepared in D2O with addition of DCl until
complete chitosan dissolution. Signals of the solvent were
suppressed by selective impulses using gradient.
First derivative ultraviolet spectrophotometry (FDUV):
Analysis was carried out according to the method described
earlier  using Shimadzu UV-1601PC spectrophotometer.
For the exclusion of the solvent influence, UV-spectra of
AcOH at different concentrations (0.01М, 0.02М, 0.03М)
were recorded and the crossing point (zero crossing point,
ZCP) was found (203nm). Calibration curve was plotted
using N-acetyl-D-glucosamine in AcOH UV spectra. Distance
from the ZCP to the spectrum (in optical density units, ΔD,
O.D.) was measured.
Asymmetric flow field fractionation: Fractionation
of polymers was carried out using AF 2000 MF Flow FFF
System (AF4) (Post nova Analytics, Germany) with following
parameters: channel thickness 350 μm, PES separating
membrane with 10 kDa cutoff limit, flow speed 0.3 ml/min.
Chitosan samples 0.1-0.5% (w/v) in 0.2 M AcOH/AcONa buffer
(pH=4.5) were prepared for the analysis. Characteristics of
each fraction were studied using refractometric (dn/dc value
was 0.181 ml/g) and multi-angle light scattering (MALS)
Dynamic viscosity studies: Determination of dynamic
viscosity (ηd) was implemented using Fungi lab Smart
L rotation viscometer under with following conditions:temperature 25°С, rotation speed 100 rpm, spindle L2.
Chitosan 1% (w/v) in 0.2M AcOH/0.1 M AcONa buffer
(pH=4.4) was prepared for the measurement. Dynamic
viscosity values (mPa.s) were calculated as an average of
three independent measurements.
Viscometry: Viscometry was conducted using three
published protocols (Table 1). Chitosan viscosity average
molecular weights (Mn
Size exclusion chromatography: Number(Mn)
average , weight average(Mw) , and z-average (Mz
) MW were determined by high-performance size exclusion
chromatography (SEC) using S2100 (Sykam, Germany)
chromatograph, Ultahydrogel-250 (7.8×300 mm) (Waters,
USA) column, GFC-4000 (4×3 mm) (Phenomenex, USA) precolumn.
Elution was carried out with 0.05М AcOH, 0.15М
AcNH4 solution (рН=5.1) at flow 0.5 ml/min and temperature
30˚C. Separation process was controlled using refractometric
detector К-2301 (Knauer, Germany). Molecular-weight
distribution values were calculated with “Multichrom” 1.6
software (“Ampersend”, Moscow, Russia). Different MW
dextrans for SEC (Sigma, USA) were used to calibrate the
a) Estimation of DD using 1H-NMR spectra: DD was
calculated from 1H-NMR spectra by two equations most
often used. In the first case DD was obtained by calculating
the ratio between proton integrals at 3.2 and 2.0 ppm (Figure
1) using equation 1.
DD= (3AH2-GlcN)/ACH3+3AH2-GlcN) .100 %, (1)
where Ax is an integral of proton X signal.
As an alternative, the equation taking into account the ratio
of integrals of all glucopyranose protons signals (Figure 1) marks
3 and 1 to the protons of acetyl group of NAcGlcN (Figure 1) was
chosen equation 2.
b) Estimation of DD using conductometric titration
curves: There are also two methods to calculate DD on the
basis of the titration curves. The first one uses only part 1 of
the curve (Figure 1b) equations 3, 4:
x = N1.V1 – N2.V2 (3)
DD= x/[x+(m.0.9-x.161)/203].100 %, (4)
where V1 is volume of acid solution, ml; V2 is volume
of alkaline solution required for the titration, ml; N1 is
normal concentration of acid solution, Eq/ml; N2 is normal
concentration of alcaline solution, Eq/ml; 161 and 203 g/mol are
the molar weights of GlcN and NAcGlcN units accordingly; m is the mass of the sample, g. The second one takes into account the
difference between volumes of NaOH used for the neutralization
of unbound HCl and consumed to titer free amino groups in the
polymer chain equation 5:
where CNaOH is NaOH concentration, mol/l; V1 and V2 is added
volumes of NaOH (ΔV), l; 161 g/mol is molar weight of GlcN unit
(C6H11O4N), m is the mass of the sample, g.
Statistical analysis was conducted using Microsoft Excel
software. Correlation analysis was carried out by Pearson’s
statistics which was shown to be suitable for any sample size
. At sample volume equal to n=3, correlation significance
р>0.95 is achieved at r > 0.997 for two-tailed test as in our case
DD is the main parameter which distinguishes chitin from
chitosan. It is defined as a molar ratio of D-glucosamine and
N-acetyl-D-glucosamine units in the polymer structure. It
affects basic physico-chemical properties of the polysaccharide
including solubility in diluted organic acids . Most studies
comparing different methods for DD determination, use
commercially available samples of chitosan with DD>70%
[3,5,22]. Chitosan with lower DD is well described in literature
however the number of studies comparing such samples using
different methods is limited [23-25]. To compare methods of
DD estimation, samples A2 and A3 with low DD were obtained
from chitosan A1 using mild conditions of acetylation preventing
decrease in chitosan MW. Samples A1-A3 were analyzed by
1Н-NMR, conductance titration (CT) , and FDUV [5,16] (Figure
1Н-NMR-spectrometry is one of the most frequently used
methods to determine DD and is recommended by the American
Standard Test Method to be used as a standard . Degradation
of high molecular weight chitosan samples before the analysis
is able to increase the quality of NMR spectra and the accuracy
of the results . However, degradation is the additional step
which may complicate the analysis. The method of degradation
can also affect DD estimation and may require sample purification
after hydrolysis. Taking into account these considerations a
degradation step was avoided. Proton spectra of A1 and its reacetylated
derivatives A2 and A3 are shown in (Figure1a). The
most specific signals are the second proton of D-glucosamine unit
(GlcN) (Figure 1) mark 1 and the acetic group proton of N-acetyl-
D-glucosamine (NAcGlcN) (Figure 1) mark 2. To determine DD
on the basis of 1H-NMR spectrum several calculation methods
are used [27,28]. DD was calculated by two equations most often
used. In the first case DD was obtained by calculation of ratio
between proton integrals at 3.2 and 2.0 ppm using equation 1 (Section 2.3). As an alternative, the equation taking into account
the ratio of integrals of all glucopyranose protons signals (Figure
1) marks 3, 1 to the protons of acetyl group of NAcGlcN (Figure
1) mark 2 was chosen equation 2.
An alternative method for DD measurement is the analysis
of the curves obtained by calculating the first derivative of UVabsorption
spectra of NAcGlcN. This method is considered to be
accurate, reproducible and does not demand high-skilled staff
and expensive equipment . FDUV curves are presented in
One more simple but accurate method to determine DD is
conductometric titration (Figure 1c). There are also two methods
to calculate DD on the basis of the titration curves equations 3-5.
The results on DD of chitosan A1-A3 obtained by different
methods are shown in (Table 1). Absolute means of DD were close
as determined by 1Н-NMR(1), FDUV, and CT(1) while 1Н-NMR(2)
and CT(2) provided lower and higher values accordingly. It
can be hypothesized that 1Н-NMR(2) decreased estimation is
a result of superimposition of glucopyranose protons signals
while overestimation provided by CT(2) is a result of chitosan
sedimentation in NaOH that impedes the detection of “plateau”
on the titration curve (Figure 1).
For the analysis of the internal relation between chitosan
samples, the data obtained by different methods were analyzed
by Pearson’s correlation statistics. For a small sample size (n=3)
non-parametric statistic is usually used. DD data analyzed
by Spearman’s rank correlation statistics provided r=±1 for
all methods used showing a high degree of correlation. More
detailed analysis of DD data by parametric Pearson’s correlation
statistics demonstrated a high positive correlation reaching 95% confidence in most cases (Table 2). For example, the coefficient
of correlation for CT (1) and CT (2) was r=1 in spite of the
overestimated means provided by CT (2) calculation. The same
high level of correlation was found between FDUV and CT (1). All
other data gave slightly lower correlations which can possibly
be explained by higher experimental errors. The estimated
errors for 1Н-NMR (1), FDUV, and CT (1) are 10 %, 2 %, and
3 % accordingly. The reliable correlation of DD data as well as
comparable DD values obtained by different methods show
that any of the methods used determines chitosan DD rather
a 1Н-NMR(1), nuclear magnetic resonance, DD is determined
using equation (1).
b 1Н-NMR(2), DD is determined using equation (2).
c FDUV, DD is determined using the first derivative of UVabsorbance
spectra of NAcGlcN.
dCT(1), conductometric titration, DD is determined using
eCT(2), conductometric titration, DD is determined using
fPearsons’s correlation coefficients with probability >95%
(r>0.997) are shown in bold.
i. Viscometry: Molecular weight (MW) of chitosan is the
second most important characteristic of the polymer that
determines its physicochemical and biological properties.
In spite of a great number of methods used to estimate MW,
many of them are secondary as they calculate MW through
hydrodynamic volume or gyration radius of the molecules
which depend on the range of different factors (nature of
solvent and its ionic strength, pressure, etc.). Many methods
use standards: proteins, dextrans, polymer beads etc., which
behavior in aqueous solutions is more predictable [29,30].
Charged polysaccharides including chitosan have a tendency to aggregate in aqueous solutions which influences their
hydrodynamic properties [7,31,32]. Methods of chitosan
MW determination include viscometry, size exclusion
chromatography (SEC), dynamic and static light scattering,
etc. Among all, viscometry is one of the simplest. Intrinsic
viscosity (η) and MW can be related by Mark-Houwink
equation: [η]=kMWα, where k and α are the empirically
determined coefficients. Empirical constants k and α for
chitosan vary significantly because of different conditions
and methods used to obtain them [4,18,19,33-39] (Table
1). As a result there are no widely accepted coefficients for
Mark-Houwink equation not only for chitosan, but also for
other charged polymers.
Dynamic viscosity is another independent and widely
used method characterizing chitosan and other polymers.
For the analysis of samples considerably differing in viscosity,
appropriate spindles are used. As with intrinsic viscosity,
buffers and temperature are also the controlled parameters. The
third standard method used to determine chitosan MW is SEC.
SEC obtains MW in three forms: number average Mn , weightaverage Mw
, and z-average Mz ; as well as it calculates dispersity
index (DI). However there are limitations for SEC. Molecular SEC
standards dextrans, pollutants, polyethylenoxide and others
are uncharged polymers. Ci S.X. et al. as well as others showed
that SEC profiles of uncharged polymers used as molecular
SEC standards, are much narrower than of charged ones such
as chitosan or alginate . This corresponds to different
physicochemical properties of charged and neutral polymers. SEC
coupled with multi-angle light scattering (MALS) detector helps
to overcome some obstacles related to the use of standards. SEC is also sensitive to the nature and ionic strength of the solvent as
they affect hydrodynamic volume of the molecules .
Field flow fractionation (FFF) is a relatively new method for
the analysis of chitosan behavior in aqueous solutions [12,29].
Contrary to SEC, fractionation by FFF occurs in flow at low
pressure and in the absence of a solid phase. Coupling of FFF with
MALS and refractometric detector gives a unique opportunity
to determine MW in different chitosan fractions. On account of
this, FFF finds a wide application for the analysis of different
polymers including natural polysaccharides such as amylase and
amyl pectin , cellulose , alginate  etc.
To compare MW determined by different methods, low MW
chitosan samples B and C (DD>90 %) were prepared. Table 1
presents the experimental conditions for viscometry and the
empirical coefficients of the Mark-Houwink equation taken
from different published protocols. The results of viscometry
experiments for samples A1-A3 are shown in (Table 3).
Application of Mark-Houwink coefficients from protocol 1 to
A1-A3 intrinsic viscosity data provided a significant increase in
MW of samples with low DD, while both intrinsic and dynamic
viscosity steadily decreased. Contrary to this, protocols 2 and
3 achieved decreasing MW with a decrease in DD, however the
absolute MW values differed significantly (t-test, p<0.01). Note
that the conditions used to obtain A2 and A3 were mild and
MW was unchanged in contrast to alkaline deacetylation which
induces partial hydrolysis of chitosan . These results support
earlier published observation that DD affects conformation
of chitosan molecule leading to a different physicochemical
behavior translated into different MW determined .
The most important observation was the stability of intrinsic
viscosity of chitosan samples determined by different protocols
(p>0.4, t-test). It should be noted that protocol 3, which used
higher salt buffer, slightly decreased the means in comparison
with protocols 1 and 2 which provided identical viscosity values.
Analysis of samples A1, B, C with high DD demonstrated that
characteristic viscosity did not vary much while absolute MW
values determined using published Mark-Hauwink coefficients
differed 3.7 to 28 times (Table 4).
ii. Determination of chitosan MW using SEC and FFF: SEC
and FFF elution profiles of chitosan samples A1-A3, B, and C
are shown in (Figure 2) and the results are summarized in
(Table 5). FFF data in (Table 5) are shown only for fraction
1 (Figure 2) which corresponds to non-aggregated chitosan.
SEC failed to fractionate accurately low DD chitosans A2 and
A3 (Table 5) possibly due to a massive chitosan aggregation
which supports the results obtained by Schatz et al. .
Both SEC and FFF showed aggregation in samples A2 and
A3 as revealed by asymmetric SEC profiles (Figure 2) and a
significant increase in fraction 2 in FFF analysis (Figure 2).
aDI - dispersity index determined by SEC or FFF accordingly
where DI= M w/M n.
Aggregation of low DD chitosan samples was shown earlier
by Popa-Nita et al. . Fractionation of chitosan samples by
FFF was undertaken to obtain chitosan samples with as low
dispersity as possible. Indeed for low MW B and C chitosan
samples dispersity index (DI) for fraction 1 in FFF was around 1
(Table 5) while for the same fraction 1 in A1 DI was found to be even higher than as determined by SEC (2.4 vs 2.0). This result
shows the intrinsic property of high molecular weight chitosan
to form aggregates. Of note, aggregation of chitosan is evident for
each sample (Figure 2) f-j, fractions 1, 2, 3.
Table 4 shows the results of Pearson’s parametric correlation
analysis of the viscosity data. In spite of a small number of samples
correlation was significant (р<0.05). At the same time there was
a high difference in MW, especially for the small molecule C (28
times). Pearson correlation coefficient is a measure of the linear distance between two variables X and Y, giving a value between
+1 and −1, where 1 is total positive correlation, 0 is no correlation,
and −1 is total negative correlation. It is widely used as a measure
of the degree of linear dependence between two variables. Thus
it means that the change in MW determined by different methods
is highly proportional. At the same time intrinsic viscosities
determined by different protocols showed very close results both
in absolute values and correlation coefficients. High correlation
between all viscosity data (r>0.999, p<0.05) (Table 4) shows that
all three protocols recognize the difference in the size of chitosan
The data on the comparison of MW measurements obtained
by viscosity, SEC, and FFF for samples A1, B, and C are presented
in (Table 6). The values of MW of each sample differed 2-8 times
however a correlation between all data was highly significant
(p<0.05). High correlation of all data obtained by different
methods demonstrates that all of them distinguish the difference
between chitosan samples. However the absolute values of MW
vary significantly possibly as a result of different physicochemical
behavior of chitosan in different conditions.
Taken together it can be concluded that each method
characterizes chitosan in a specific condition and none of them
can produce the same MW values for the same sample, which is
likely to mean than chitosan changes its conformation in different
conditions and behaves as a molecule with different MW. The
molar concentration of the buffers used for the viscometry
studies increased in a row: 0.3<0.4<0.5M for protocols 1-3
accordingly. MW of the same samples determined by these
protocols decreased with the increase in molar concentration
(Table 4). At the same time viscosity did not change much (Table 4). Data on the decrease of both viscosity and MW in A1
after reacetylation (chitosan samples A2 and A3) support this
conclusion. Viscometry, as the simplest and most direct method
of chitosan analysis, should be preferred. Chitosan viscosity,
either intrinsic or dynamic, should serve as a main chitosan
Problems of MW determination are not unique for chitosan.
Contrary to DNA and proteins, which have stable conformation,
polysaccharides are prone to change it in different conditions
(salt, temperature, pressure) and can easily form complexes [46-
50]. Our results demonstrate that MW may not be considered as
the best characteristic of chitosan. Contrary to MW, viscosity is
a much more stable characteristic of chitosan as well as of other
The results of this paper demonstrated that deacetylation
degree of chitosan molecules is a stable characteristic and can
be determined sufficiently well by multiple modern methods. Contrary to DD, we and many others showed that MW severely
depends on the method of determination which usually is
explained as a result of methodological problems: recalculation
of MW from hydrodynamic volume or gyration radius using
empirical coefficients; usage of neutral standards to obtain MW
by SEC; different input of aggregation in some methods et cetera.
At the same time, in spite of differences in the means of chitosan
MW, Pearson’s coefficients of correlations demonstrated
high relation (r=1) between all the methods used (p<0.05)
showing that all of them are sensitive to the changes in chitosan
molecule mass or deacetylation degree. Taken collectively we
have hypothesized that chitosan changes its conformation in
different physicochemical conditions and behaves as a molecule
with different MW. Contrary to many other methods, including
highly advanced such as FFF, viscosity characterizes chitosan
directly without recalculations and standards; it can be used for
a wide range of chitosan molecules including samples with low
MW and DD which are difficult for the analysis. We showed that
viscosity data perfectly correlate with SEC and FFF data, while
MW obtained by recalculation using Mark-Houwink coefficients,
different MW standards, or gyration radius differ significantly.
The uncertainty of chitosan MW determination is one of the
main obstacles to its medicinal application. It can be suggested
that the main characteristics of chitosan are DD and viscosity
which completely and reliably identify the nature and predict the
behavior of chitosan used for biomedical applications.
The research was supported by the Russian Scientific
Foundation, grant №16-14-00046. We are grateful to Post nova
Analytics Company (Germany) for FFF analysis of chitosan
samples; to Dr. Valery Varlamov (Institute of Bioengineering,
Moscow, Russia) for the fruitful discussion of the results; to
Dr. Maxim Dubinnyi (IBCh, RAS, Moscow, Russia) for 1H-NMR