Determination of condensed Tannin Monomers in environmental samples by capillary gas chromatography of acid depolymerization extracts

Article abstract of DOI:10.1021/ac991301y

A method for molecular-level quantification of condensed tannin is described that uses acid depolymerization and carbocation capture by phloroglucinol. Resulting monomers and phloroglucinol adducts are trimethylsilyl derivatized, separated by capillary gas chromatography, detected by flame ionization, and quantified relative to standards. Optimal depolymerization conditions were determined for acid strength, phloroglucinol concentration, time, and temperature. The method gives reproducible results in leaf litter that are linear over 2 orders of magnitude with detection limits down to ～100 ng condensed tannin. In addition to tannin, triterpenoids were also identified and can be quantified with this method. Analyses of soils and sediments indicate that mineral interactions with condensed tannin are important both analytically and environmentally.

Full text of DOI:10.1021/ac991301y

Anal. Chem. 2000, 72, 5115-5124
Technical Notes
Determination of Condensed Tannin Monomers in
Environmental Samples by Capillary Gas
Chromatography of Acid Depolymerization
Extracts
Peter J. Hernes*,† and John I. Hedges
School of Oceanography, University of Washington, Seattle, Washington 98195-7940
elsewhere.3-6
A method for molecular-level quantification of condensed
tannin is described that uses acid depolymerization and
carbocation capture by phloroglucinol. Resulting mono-
mers and phloroglucinol adducts are trimethylsilyl de-
rivatized, separated by capillary gas chromatography,
detected by flame ionization, and quantified relative to
standards. Optimal depolymerization conditions were
determined for acid strength, phloroglucinol concentra-
tion, time, and temperature. The method gives reproduc-
ible results in leaf litter that are linear over 2 orders of
magnitude with detection limits down to ∼1 0 0 ng con-
densed tannin. In addition to tannin, triterpenoids were
also identified and can be quantified with this method.
Analyses of soils and sediments indicate that mineral
interactions with condensed tannin are important both
analytically and environmentally.
Tannin in the environment is of interest because it is quanti-
tatively important and participates in a number of important
transformations, including photochemical and redox reactions,
nitrogen immobilization, and cation complexation. Tannin is also
a potential precursor to humic substances via quinone formation
and subsequent condensation reactions with proteins and amino
acids. Although a number of depolymerization methods for tannin
can be found in the natural products literature, application to
environmental samples has been limited almost exclusively to bulk
techniques. Because of the potential wealth of information at the
molecular level, the aim of this study was to develop a molecular
technique suitable for environmental samples. Further objectives
were that the technique should utilize commonly available
instrumentation, require minimal chemical manipulations or
hazardous reagents, and should allow rapid throughput.
Existing methodologies for tannin analyses contain many
drawbacks either inherent in the method7,8 or when applied to
environmental samples.7-9 Briefly, these include lack of appropri-
ate standards, analyses of extracts instead of whole samples,
unpleasant or hazardous reagents, lack of reproducibility or
quantification, and complicated chromatography or less common
instrumentation. Among the molecular techniques for condensed
tannin, while nearly all utilize acid depolymerization with capture
of the carbocation by a nucleophile to form an adduct (Figure 2),
there is a lack of standardization among depolymerization condi-
tions. The various techniques utilize depolymerization times from
Tannin is the fourth most abundant biochemical in terrestrial
biomass, following cellulose, hemicellulose, and lignin. In rapidly
cycling soft tissues such as leaves and needles, however, tannin
is often more abundant than lignin1,2 and, therefore, an important
component of the carbon cycle. Tannin consists of two types,
condensed and hydrolyzable tannin (Figure 1). The former exists
as oligomers and polymers of three-ring flavanols (note the
terminology of extender units and terminal units in Figure 1),
while the latter is made up of gallic acid or its derivatives often
esterified to polyols such as glucose. The most common linkages
in condensed tannin are 4f8, while 4f6 linkages lead to
branching (Figure 1). This type of condensed tannin is also
referred to as proanthocyanidin because of its tendency to form
cyanidins or related compounds upon acid depolymerization.
Other linkages in condensed tannin include A- to A-ring and A-
to B-ring, which, strictly speaking, are not proanthocyanidins.
More complete reviews of structure and chemistry can be found
(3) Hemingway, R. W. In Chemistry and Significance of Condensed Tannins;
Hemingway, R. W., Karchesy, J. J., Eds.; Plenum Press: New York, 1988;
pp 83-107.
(4) McGraw, G. W. In Chemistry and Significance of Condensed Tannins;
Hemingway, R. W., Karchesy, J. J., Eds.; Plenum Press: New York, 1988;
pp 227-248.
(5) Laks, P. E. In Chemistry and Significance of Condensed Tannins; Hemingway,
R. W., Karchesy, J. J., Eds.; Plenum Press: New York, 1988; pp 249-263.
(6) Hemingway, R. W. In Chemistry and Significance of Condensed Tannins;
Hemingway, R. W., Karchesy, J. J., Eds.; Plenum Press: New York, 1988;
pp 265-283.
† Present address: Baruch Institute, University of South Carolina, Columbia,
SC 29208.
(1) Hedges, J. I.; Weliky, K. Geochim. Cosmochim. Acta 1 9 8 9 , 53, 2659-2673.
(2) Benner, R.; Hatcher, P. G.; Hedges, J. I. Geochim. Cosmochim. Acta 1 9 9 0 ,
54, 1991-2001.
(7) Mole, S.; Waterman, P. G. Oecologia 1 9 8 7 , 72, 137-147.
(8) Hagerman, A. E.; Butler, L. G. J. Chem. Ecol. 1 9 8 9 , 15, 1795-1810.
(9) Appel, H. M. J. Chem. Ecol. 1 9 9 3 , 19, 1521-1551.
10.1021/ac991301y CCC: $19.00 © 2000 American Chemical Society
Published on Web 09/12/2000
Analytical Chemistry, Vol. 72, No. 20, October 15, 2000 5115

Figure 1. Structures of typical condensed and hydrolyzable tannins.
15 min10 to 50 h,11 temperatures from room temperature12,13 to
>100 °C,14 acid strengths from 0.02 10 to 2.5 M,15 and nucleophile
concentrations from 0.04 12 to 2.4 M.16 Several different acids have
been tried, including acetic, sulfurous, phosphoric, hydrochloric,
and several carboxylic acids. The nucleophile predominantly
consists of benzyl mercaptan or phloroglucinol. Solvents include
ethanol, methanol and water, and dioxane and water.
Various chromatographic techniques have been utilized for
separation of depolymerization mixtures, including paper chro-
matography, column chromatography, liquid chromatography (see
ref 17 for review), and gas chromatography (GC).10,14 The products
are often methylated prior to chromatography, which involves
either explosive or toxic reagents. Less common for GC are the
formation of trimethylsilyl (TMS) derivatives, which avoid the
safety hazards of methylation but increase molecular weight
significantly. Nevertheless, with today’s high-temperature gas
chromatographic columns and electronic pressure controllers,
increased molecular weight is less of an issue.
above, but using acetone/ water (70:30 v/ v) as the solvent.
Acetone/ water is a commonly used extraction solvent for obtain-
ing purified condensed tannins prior to depolymerization in
most techniques7 and was a natural choice since we desired a
method appropriate for analyzing whole samples. Although benzyl
mercaptan often gives higher yields of adducts than phloroglu-
cinol,14 it is also unpleasant to work with and so we opted for the
latter. For the acid, we chose hydrochloric because its volatility
allows easy removal after depolymerization if so desired. Finally,
we chose TMS derivatization and capillary gas chromatography
coupled with a flame ionization detector (FID) for detection. TMS
derivatization is easy, the derivatives are stable, and GCs offer
relatively high resolution and ready mass spectral analysis.
A primary reason for analyzing whole samples instead of tannin
extracts is the potential for measuring non-tannin compounds
within our analytical window. Two compound classes that can be
measured by this technique include carboxylic acids and triter-
penoids. The former include short-chain carboxylic acids as well
as some cutin acids. The latter are routinely quantified and include
unique terrestrial biomarkers.18 Together with tannin, they offer
a powerful tool for geochemical studies.
The analytical procedure described here was developed by
utilizing many of the depolymerization conditions referred to
(10) Rigaud, J.; Peres-Ilzarbe, J.; Ricardo Da Silva, J. M.; Cheynier, V. J.
Chromatogr. 1 9 9 1 , 540, 401-405.
(11) Matsuo, T.; Tamaru, K.; Itoo, S. Agric. Biol. Chem. 1 9 8 4 , 48, 1199-1204.
(12) Koupai-Abyazani, M. R.; McCallum, J.; Bohm, B. A. J. Chromatogr. 1 9 9 2 ,
594, 117-123.
(13) Foo, L. Y.; Karchesy, J. J. Phytochemistry 1 9 8 9 , 28, 3185-3190.
(14) Matthews, S.; Mila, I.; Scalbert, A.; Pollet, B.; Lapierre, C.; Herve du Penhoat,
C. L. M.; Rolando, C.; Donnelly, D. M. X. J. Agric. Food Chem. 1 9 9 7 , 45,
1195-1201.
(15) Brown, B. R.; Shaw, M. R. J. Chem. Soc., Perkin Trans. 1 1 9 7 4 , 2036-
2049.
(16) Thompson, R. S.; Jacques, D.; Haslam, E.; Tanner, R. J. N. J. Chem. Soc.,
Perkin Trans. 1 1 9 7 2 , 1387-1399.
EXPERIMENTAL SECTION
Sample P reparation. Soils, sediments, and leaf samples are
freeze-dried and ground to pass through a 0.351-mm sieve. Green
leaves from Camellia sp. and dripline soil were collected in Seattle,
WA. Sediments were collected from Dabob Bay, WA. Whole
brown leaves were collected from the litter of pin oaks (Quercus
palustris) on the University of Washington campus in Seattle, WA.
Condensed Tannin Depolymerization. Approximately 50 mg
of sample and 100 mg of phloroglucinol is weighed into a 50-mL
(17) Karchesy, J. J.; Yougsoo, B.; Chalker-Scott, L.; Helm, R. F.; Foo, L. Y. In
Chemistry and Significance of Condensed Tannins; Hemingway, R. W.,
Karchesy, J. J., Eds.; Plenum Press: New York, 1988; pp 139-151.
(18) Hernes, P. J.; Hedges, J. I.,submitted to Geochim. Cosmochim. Acta.
5116 Analytical Chemistry, Vol. 72, No. 20, October 15, 2000

Figure 2. Acid depolymerization of a condensed tannin dimer with and without phloroglucinol. Structures of all compounds quantified are
shown. Numbers in parentheses correspond to DB5ms elution order as shown in Figure 3a.
Pyrex centrifuge tube (12-24 samples at time depending on
heating bath capacity) with a Teflon-lined phenolic screw cap. After
addition of a Teflon-coated magnetic stir bar, 0.9 mL of 0.333 M
HCl and 2.5 mL of acetone (both Ar-sparged for 10 min) are added
to each tube. Each tube is purged with Ar (thereby evaporating
∼0.4 mL of acetone for a final total volume of ∼3 mL) and capped
before being placed in a constant-temperature bath (30 °C)
equipped with a magnetic stirrer. Samples are depolymerized for
24 h.
Sample Extraction and Cleanup. Upon removal from the
constant-temperature bath, internal standard (predominantly ∼100
µg of hematoxylin in 20 µL of pyridine in this study) is added
along with 10 mL of deionized water. Approximately 4 mL of ethyl
acetate is added to each tube. Tubes are shaken for 30+ s and
centrifuged briefly, and then the ethyl acetate fraction is drawn
off the top with a Pasteur pipet and dripped through anhydrous
NaSO4 columns to remove water. The columns drain into 1-dr
vials under a stream of N2. After three extractions, a two-thirds
split is removed from each sample and placed in another 1-dr vial.
All samples are taken to near dryness and then placed in a
vaccuum desiccator overnight to remove any residual water, HCl,
or solvent.
Derivatization and Gas Chromatography. Samples (one-
third splits-the remainder was archived for future analyses) are
redissolved in 200 µL of pyridine, and an aliquot (20-30 µL) is
placed in an autosample vial containing a 200-µL insert. An equal
volume of Regisil ((bis(trimethylsilyl)trifluoroacetamide) + 1%
trimethylchlorosilane; Regis Chemical Co.) is added to each vial,
Analytical Chemistry, Vol. 72, No. 20, October 15, 2000 5117

Figure 3. Gas chromatographic traces on (a) DB5ms and (b) DB35ms liquid phases of the trimethylsilyl derivatives of tannin and triterpenoid
compounds obtained by acid depolymerization of pin oak leaf litter using the standard depolymerization conditions. Gas chromatographic equipment
and conditions are described in the text. Identifications: 1, hematoxylin (internal standard); 2, epicatechin; 3, catechin; 4, epigallocatechin; 5,
stigmast-5-en-3â-ol; 6, â-amyrin; 7, epicatechin (4âf2) phloroglucinol; 8, R-amyrin; 9, catechin (4âf2) phloroglucinol; 10, oleanolic acid; 11,
catechin (4Rf2) phloroglucinol; 12, ursolic acid; 13, epigallocatechin (4âf2) phloroglucinol; 14, gallocatechin (4Rf2) phloroglucinol.
the vial is capped, the contents are mixed with a syringe, and the
vials are placed in a 60 °C heating block for 15 min.
loading, the oven temperature is ramped up to 200 °C at 25 °C/
min. The temperature ramp is then decreased to 4 °C/ min until
330 °C, at which point the temperature is held until the end of
the run, usually 20-50 min. If the gas chromatograph is equipped
with an electronic pressure controller (EPC), the initial column
head pressure used for the DB5ms is 20 psi. After 9.2 min (i.e.,
when the oven reaches 250 °C), the pressure is increased 0.5 psi/
min to 35 psi and held until the end of the run. For the DB35ms,
initial pressure is held at 13 psi for 2 min, then ramped at 1 psi/
min to 30 psi, and maintained until the end of the run. A DB5ms
chromatogram of TMS-treated hydrolysate from pin oak leaf litter
can be found in Figure 3a, while the same sample injected on a
DB35ms is shown in Figure 3b.
A Hewlett-Packard 5890 gas chromatograph fitted with an FID
or interfaced to a Hewlett-Packard 5970 mass-selective detector
is used for all analyses. The injection port and detector are
maintained at 310 °C. The injection port is set up for “splitless”
injection using a 2-mm gooseneck liner. Analyses are made by
using 30 m by 0.25 mm i.d. fused-silica capillary columns, coated
with either DB35ms or DB5ms liquid phase (J&W Scientific Inc.).
After initial injection with the oven temperature held at 70 °C,
the purge valve is kept in the closed position to allow column
loading. After 2 min, the purge valve is opened to flush any
remaining organic vapors out of the injection port. For the DB5ms,
the oven temperature is ramped up to 250 °C at 25 °C/ min. The
temperature ramp is then decreased to 3 °C/ min until 295 °C.
Finally, the temperature ramp is decreased to 1.5 °C/ min until
325 °C, at which point the temperature is held until the end of
the run, usually 20-30 min. For the DB35ms, following column
P eak Identification and Quantification. Most peak identities
were determined by comparing mass spectra to those of available
authentic standards (epicatechin, catechin, epigallocatechin,
R-amyrin, oleanolic acid, ursolic acid) and to those in a library of
trimethylsilyl-derivatized natural compounds generated by W.
5118 Analytical Chemistry, Vol. 72, No. 20, October 15, 2000

Table 1. Diagnostic Mass Spectral Fragments (m/z) of Trimethylsilyl Derivativesa
Tannin
catechin/ epicatechin (PC)
afzelechin/ epiafzelechin (PP)
gallocatechin/ epigallocatechin (PD)
gallocatechin/ epigallocatechin (PD)
acetone derivative
epicatechin 4â adduct
catechin 4R and PC adduct
PP adduct
368 (B-ring), 355 (A-ring), M+ 650
280 (B-ring), 355 (A-ring), M+ 562
456 (B-ring), 355 (A-ring), M+ 738
409/ 424 (B-ring), 355 (A-ring), M+ 706
368 (B-ring), 533 (A-ring + adduct), 648
368 (B-ring), 533 (A-ring + adduct)
280 (B-ring), 533 (A-ring + adduct)
456 (B-ring), 533 (A-ring + adduct)
409/ 424 (B-ring), 533 (A-ring+adduct)
PD adduct
PD adduct acetone derivative
Triterpenoid
R-amyrin
â-amyrin
oleanolic acid
ursolic acid
218, 189, 203, M+ 498
218, 203, 189, M+ 498
203, 320, 482, 189, 585, M+ 600
203, 320, 482, 189, 585, M+ 600
a
Abbreviations: PC, procyanidin; PP, propelargonidin; PD, prodelphinidin. See Figure 2 for structures.
Greenaway and colleagues (stigmast-5-en-3â-ol, â-amyrin). These
compound structures are shown in Figure 2. Phloroglucinol
adducts were identified by mass spectral fragmentation patterns
relative to unadducted parent monomers, relative peak areas,
elution orders, and depolymerization of a catechin (4Rf8)
catechin dimer (also designated as B-3, acquired from Leuven
Bioproducts) standard in the presence of phloroglucinol. Each
extender unit when hydrolyzed to a carbocation can form two
adducts at C-4 (see extender unit ACB in Figure 2), which IUPAC
rules designate as R and â (â as shown in Figure 2 is “up” from
the plane of the molecule as typically drawn). Thus, the isomeric
pair catechin/ epicatechin, or procyanidin (PC), gives rise to four
adducts, as does gallocatechin/ epigallocatechin, or prodelphinidin
(PD). Mass spectral fragmentation patterns for the four adducts
in each pair are nearly identical. Steric hindrance from the C-3
hydroxyl group is the primary factor that determines the orienta-
tion of the adduct, although stereochemistry of the B-ring at C-2
(which is almost always R) will also be a factor. When the C-3
hydroxyl group and the C-2 B-ring are cis to each other (as in all
the epi forms), then the â position should be highly favored by
the adduct. If the C-3 hydroxyl group and C-2 B-ring are trans to
each other, then a mixture of R and â adducts is possible. In this
study, three PC adducts and two PD adducts were identified from
mass spectral fragmentation patterns. Two PC adducts, catechin
(4âf2) phloroglucinol and catechin (4Rf2) phloroglucinol, were
identified from depolymerization products of the catechin (4Rf8)
catechin dimer. The third PC, then, is epicatechin (4âf2)
phloroglucinol. The elution order as determined for the PC
adducts is applicable to the PD adducts as well, which were
identified as gallocatechin (4Rf2) phloroglucinol and epigallo-
catechin (4âf2) phloroglucinol. Primary mass spectral fragments
for all routinely quantified compounds are given in Table 1.
Identified compounds for which commercial standards are un-
available were quantified using the area response of the internal
standard.
an extracted and purified condensed tannin from the same leaves.
The following routine depolymerization conditions were estab-
lished: 1.0 M HCl in acetone/ water (70:30 v/ v), 0.26 M phloro-
glucinol, 30 °C for 24 h. Other reported depolymerization
conditions when phloroglucinol was used include 0.2 M HCl in
dioxane/ water (1:1 v/ v), 0.13 M phloroglucinol, 80 °C for 20 min;14
0.3 M HCl in ethanol, 0.04 M phloroglucinol, ambient temperature
overnight;12 and 0.3 M HCl in ethanol, ∼0.7 M phloroglucinol,
ambient temperature for 30 min.13 The basis for how these
conditions were chosen was not reported. In our preliminary
experiments, yield efficiencies from extracts were noticeably lower
than the whole samples (∼5% vs an estimated 20-50%), which
raised concerns about chemical alteration during extraction and
the overall suitability of extracts as proxies for environmental
samples. Subsequently, the emphasis was shifted toward whole
samples, since the ability to analyze the latter was a primary
objective of this study. Depolymerization series were conducted
by holding three of the conditions constant and varying the fourth.
All analyses were done in duplicate, and the results presented
are averages with error bars indicating the range. The acid
strength series was carried out on both green camellia leaf
material (CLM) and pin oak leaf litter (POLL). Only POLL was
used for the remaining experiments for a variety of reasons,
including the presence of both PC and PD tannin, the presence
of triterpenoids that also fall in our analytical window, and the
fact that litter more closely resembles the degraded environmental
samples that we ultimately wish to analyze.
Acid Strength. Acid concentrations were varied from 0 to 2
M HCl for both CLM and POLL. On the basis of the 0 M HCl
results, CLM contained nearly 2 wt % catechin + epicatechin as
monomers (Figure 4a) while POLL contained no tannin in the
monomeric form (Figure 4b). Catechin and epicatechin were
present as both terminal units and extender units in CLM, while
POLL contained catechin, epicatechin, gallocatechin, and epigal-
locatechin terminal and extender units. Maximum tannin yields
from CLM were achieved at 0.2-0.5 M HCl, while for POLL,
maximum yields were achieved at 1.0 M HCl. Considering the
degradation resistance needed for tannin to persist into the litter
stage, the higher acid strength necessary for POLL is not
RESULTS AND DISCUSSION
A primary issue in method development is the choice of
representative sample material. Preliminary experiments were
conducted using whole green leaves from Camellia sp. as well as
Analytical Chemistry, Vol. 72, No. 20, October 15, 2000 5119

Figure 4. Depolymerization of (a) CLM and (b) POLL with varying acid strengths. Note that, in order to plot on the same scale, in (b) and all
following plots, degree of polymerization was divided by 2.5 and triterpenoids were divided by 5.
surprising. The fact that tannin yields decrease at higher acid
strengths in both sample types (especially monomeric catechin
in CLM) indicates a competing degradation reaction. A further
problem with higher acid concentrations is the increased yield of
phloroglucinol reaction products that can potentially coelute with
peaks of interest. The primary interference occurs with the
DB35ms column between monomeric catechin and the peak
immediately preceding it in Figure 3b. Below 1.0 M HCl, catechin
can be resolved cleanly on the DB35ms column. At 1.5 and 2 M
HCl, the phloroglucinol reaction product peak was larger than
catechin and made resolution difficult. These coelution problems
did not occur on the DB5ms, which proved to be a better overall
column choice for POLL. Given the complexity of the chromato-
grams resulting from this technique in general, dual-column
analyses may be appropriate.
Only two other studies were found that tested acid concentra-
tions for phloroglucinol depolymerization. Matsuo et al. in 198411
found that kaki-tannin yields increased with increasing acetic acid
concentration, but stopped at 2.5 M before a plateau was reached.
Matthews et al. in 199714 tested 0.05 and 0.1 M HCl at ambient
temperatures and 80 °C and found that the higher acid concentra-
tion resulted in a 20-25% yield increase. Given the increases seen
in this study at up to 1.0 M HCl, it would be informative to see
the experiments of Matthews et al. repeated at higher HCl
concentrations.
extender units by acid depolymerization. At higher acid strengths,
it appears that catechin and epicatechin are more susceptible to
degradation during acid depolymerization than the adducted
extender units. This may be because they are, in fact, more readily
degraded. Kinetics also likely come into play, however, as nearly
80% of the catechin is present in CLM as a monomer and
immediately available for degradation, whereas the extender units
are released over time.
Calculated degree of polymerization in POLL shows a trend
similar to CLM, initially increasing with acid strengths up to 1.0
M HCl before exhibiting a slight decrease at 1.5 M (Figure 4b).
At 2.0 M HCl, no terminal units were detected; therefore, a degree
of polymerization could not be calculated. As with CLM, though,
the adducted extender units persist. Because both terminal and
extender units are released at the same time, this appears to
indicate that adduction imparts increased stability.
Finally, the sum of the triterpenoids â-amyrin, R-amyrin,
oleanolic acid, and ursolic acid show little trend (Figure 4b. Note
division by 5 to plot on the same scale). They do not appear to be
covalently bound to the plant matrix, given that the yield at 0 M
HCl is similar to the yields when acid is present. Extra care must
be taken when quantifying the triterpenoids given the chromato-
graphic conditions used in this method. In particular, response
factors relative to the internal standard can vary considerably when
column overloading occurs with the triterpenoids, and the amount
of standards injected on the GC must be matched more carefully
to the amount present in the sample injected.
In addition to individual tannin compounds, a calculated degree
of polymerization (extender plus terminal units as measured by
this technique divided by terminal unitssan indication of chain
length) is shown for CLM in Figure 4a. Because shorter polymers
are generally more labile than longer polymers, degree of
polymerization is a potential diagnostic parameter for diagenesis
in environmental samples. In CLM, the calculated degree of
polymerization increased with increasing acid strength. At 0.2 and
0.5 M HCl, this is simply indicative of the higher release of
Because it gave the highest yields in POLL, 1.0 M HCl was
used for the remainder of the experiments.
Depolymerization Time. POLL was depolymerized from 1
to 48 h. Nearly half the total yield was achieved within the first
hour, and a plateau was reached by 24 h (Figure 5). This pattern
shows the reactivity of condensed tannin to relatively mild acid
depolymerization and indicates that the reaction products are
5120 Analytical Chemistry, Vol. 72, No. 20, October 15, 2000

temperature conditions appear to decrease yields. Again, terminal
units are less stable than adducted extender units, as evidenced
by the large decrease in catechin and the large increase in the
calculated degree of polymerization (3.4-5.9) under harsher
conditions. Higher temperatures also lead to larger numbers and
quantities of phloroglucinol reaction products, which resulted in
coelution problems.
As mentioned previously, Matthews et al.14 conducted experi-
ments at ambient temperatures and at 80 °C. While depolymeri-
zation times were very different (>200 h vs 20 min.), total yields
were similar.
Although steric hindrance is the primary factor in determining
relative proportions of R and â adducts, temperature can also likely
play a role. A case in point appears to be the formation of PD
adducts. At 22 and 30 °C, only two PD adducts (epigallocatechin
(4âf2) phloroglucinol and gallocatechin (4Rf2) phloroglucinol)
give quantifiable peaks. A third PD adduct is detectable by mass
spectroscopy in trace amounts. However, at 40 °C, a distinct peak
is evident (likely gallocatechin (4âf2) phloroglucinol), and at 50
°C, it is of the same magnitude as gallocatechin (4Rf2) phloro-
glucinol.
In many studies, ambient temperature is used for depolymer-
ization when phloroglucinol is used as the nucleophile. However,
to reduce variability, a higher controllable temperature is desirable
if it does not sacrifice yields. And in fact, 30 °C gave higher yields
and so was chosen.
Nucleophile Concentration. Phloroglucinol concentrations
were varied from 0.066 to 0.53 M (as compared to the 0.04 to
∼0.7 M range in the studies cited above). For 50 mg of POLL,
this is stoichiometrically in excess of the condensed tannin
extender units by a factor of 250-2000. Tannin adduct yields
increased up to 0.26 M phloroglucinol (>20% higher than 0.066
M, but within 5% of 0.13 and 0.195sagain with a replicability of
∼5%) before tailing slightly (∼10% decrease). The tailing may be
in part related to the noticeable viscosity increase observed in
the samples with higher phloroglucinol concentrations throughout
sample workup. For instance, catechin yields should be unaffected
by nucleophile concentrations and yet also exhibited a ∼10%
decrease. Redissolving the one-third split from a 0.26 M phloro-
glucinol depolymerization sample in less than 200 µL of pyridine
often leads to derivatization problems. Thus, the 0.39 and 0.53 M
samples required a larger volume of pyridine and introduced
variability into the procedure.
The calculated degree of polymerization increased with in-
creasing phloroglucinol. This is as expected, given that an
increased nucleophile concentration should lead to more efficient
capture of the extender carbocations. Because adduct yields
decrease at the highest phloroglucinol concentrations, obviously
capture efficiency is not the only factor.
Choosing the best nucleophile concentration is perhaps the
least obvious of the four depolymerization conditions varied; 0.13,
0.195, and 0.26 M all appear to be reasonable choices. The
concentration 0.26 M was adopted because it gave the highest
yield of epicatechin (4âf2) phlorolglucinol. Epicatechin extender
units from which it is derived are the most abundant condensed
tannin.19,20
Figure 5. Depolymerization of POLL with varying depolymerization
times.
stable. Since adduction of the carbocation in theory is reversible,
while decomposition of the carbocation to a cyanidin or delphinidin
is not, yields might have been expected to decrease with time. At
ambient temperatures, Matthews et al.14 found that >200 h was
needed to achieve a plateau with their depolymerization condi-
tions. Again, our acid strength was a factor of 10 greater, which
likely explains the difference.
The calculated degree of polymerization also reaches a plateau
at 24 h (Figure 5). The initial value is 2.4 increasing to 3.5. If all
interflavan bonds were equally accessible or hydrolyzable, then
this ratio should not change. The observed increase suggests
that, in this sample, terminal units on average are more easily
cleaved than the extender unit. It is possible that the ends of
the polymer are equally cleaved (one terminal and one extender
unit) and that depolymerization of internal extender units is
inhibited until the ends are released. Since the actual conforma-
tion of condensed tannin polymers is not known, it is not clear
whether the difference in rate of depolymerization is due to
hindrance by the tannin itself or by compounds associated with
tannin.
On the basis of the plateaus in yields and calculated degree of
polymerization, 24 h is an appropriate choice for depolymerization
times and was used in the remainder of the experiments.
Depolymerization Temperature. Temperatures for depo-
lymerization were varied in ∼10 °C steps from room temperature
(22 °C) to 50 °C. The highest adduct yields were achieved at 30
°C: 20-30% higher than at 22 and 50 °C for the major adducts
and ∼10% higher than at 40 °C, as compared with the average
reproducibility of ∼5%. As with increased acid strength, harsher
(19) Czochanska, Z.; Foo, L. Y.; Newman, R. H.; Porter, L. J. J. Chem. Soc., Perkin
Trans. 1 1 9 8 0 , 2278-2286.
Analytical Chemistry, Vol. 72, No. 20, October 15, 2000 5121

if the recovery of tetrahydroxybenzophenone is assigned 1.00,
recovery of trihydroxybenzophenone was 1.14 (i.e., 14% more was
recovered in the ethyl acetate phase than tetrahydroxybenzophe-
none), and hematoxylin with five hydroxyl groups was 0.93 (i.e.,
7% less), showing the predicted trend. On this scale, most
hydroxylated flavones, flavanones, and flavanols also range from
0.9 to 1.1. Triterpenoids and unhydroxylated flavone and flavanone
ranged from 1.2 to 1.4. If the amount of water added after
depolymerization was reduced, all recoveries would converge
toward 1.0. The tradeoff in this case would be more complex
chromatography. In terms of partitioning, all three internal
standards appear to be reasonable choices. A second factor in
recovery, however, is reactivity with the sample, and as will be
shown below, hematoxylin is problematic when mineral matrixes
are involved.
Table 2. Depolymerization-Yield Linearity Testsa
1-50-mg samples 1-100-mg samples
r2
wt %
r2
wt %
Tannin
catechin
epicatechin
0.997
0.663
0.980
0.998
0.994
0.994
0.989
0.963
0.087
0.014
0.020
0.171
0.013
0.032
0.177
0.053
0.997
0.895
0.989
0.988
0.994
0.993
0.997
0.990
0.080
0.014
0.021
0.143
0.012
0.030
0.170
0.056
epigallocatechin
epicatechin 4â adduct
catechin 4â adduct
catechin 4R adduct
epigallocat 4â adduct
gallocat 4R adduct
Triterpenoid
R-amyrin
0.999
0.998
0.992
0.988
0.182
0.991
0.996
0.991
0.994
0.156
0.118
0.400
0.135
â-amyrin
0.130
0.461
0.147
oleanolic acid
ursolic acid
Chromatographically, all three internal standards elute before
any of the compounds of interest. Ideally, a late-eluting internal
standard could also be found. Because of a partial coelution
problem with trihydroxybenzophenone on the DB35ms column,
hematoxylin was used exclusively in the depolymerization series
of this study. With the incorporation of the DB5ms, however,
coelution does not appear to be a problem with the benzophe-
nones, and therefore, either is an appropriate choice.
As stated earlier, acetone/ water (70:30 v/ v) was chosen as the
depolymerization solvent because of its common use in extracting
tannin. An unexpected result of this is the quantitative conversion
of PD tannins in a reaction involving acetone, the B-ring, and one
of the hydroxyl groups. One other study observed this reaction
when irradiating catechin and CBr4 (1:1 molar ratio) in moist
acetone.21 Those investigators determined that the acetone bonds
with C-2 on the B-ring and the C-3 hydroxyl group to form a fourth
ring (Figure 2). The effect of this on mass spectra is the addition
of 56 mass units with a readily lost methyl group. Thus, instead
of a 456 fragment analogous to the 368 fragment of PC tannin
and the 280 fragment for propelargonidin (PP) tannin, what is
actually seen is a 409/ 424 doublet (Table 1). Based on experi-
ments with epigallocatechin standard, the rate of reaction is on
the order of minutes, the conversion quantitative, and the product
stable at least through a 24-h depolymerization.
Matrix Effects. Matrix effects were evaluated by the addition
of 100-1000 mg of sediment from Dabob Bay, WA, to ∼50 mg of
POLL, or ∼0.3 mg of tannin. A single addition of 1000 mg of soil
to 50 mg of POLL was also done for comparison. The effective
tannin concentration relative to the sediment, then, was 0.3-0.03
wt % tannin and 0.03 wt % relative to the soil. Because hematoxylin
proved to be very reactive toward the sediment and soil, quanti-
tation was carried out by normalizing to the sum of the two
amyrins, oleanolic and ursolic acid, since the latter were shown
to be relatively unchanged in various conditions. Not surprisingly,
the PD tannins showed the highest reactivity with the mineral
matrixes, decreasing by more than a factor of 2 with the addition
of 100 mg of sediment. All PC tannins remained unchanged. At
250 mg of sediment, all tannin showed considerable decreases,
again by a factor of 2 as compared to the 100-mg addition. In
contrast to harsher depolymerization conditions in the earlier
experiments, the calculated degree of polymerization decreased
a
Depolymerization-yield linearity for pin oak leaf litter (duplicates
of 1, 5, 10, 25, 50, and 100 mg samples, 45.1% OC). Correlation
coefficients (r2) calculated by linear regression.
Linearity. Depolymerization-yield linearity tests were carried
out on POLL, the results of which are presented as correlation
coefficients (r2) in Table 2. Sample sizes ranged over 2 orders of
magnitude: 1, 5, 10, 25, 50, and 100 mg. When the entire range
is included, all correlation coefficients exceeded 0.98, with the
exception of epicatechin at 0.90. When plotted, however, com-
pounds from the 100-mg samples all fall below the linear regres-
sion line. If linear regressions are calculated for only the 1-50-
mg samples, correlation coefficients are similar to those obtained
for 1-100 mg, but sample weight percents increase by as much
as 15%. Based on the weight percent of the lowest two compounds,
epicatechin and catechin (4âf2) phloroglucinol, and the 1-mg
sample size, GC detection limits for this technique are at least as
low as 130 ng of tannin.
Internal Standards and Standard Recoveries. Acquisition
of appropriate standards is challenging. Catechin, epicatechin, and
epigallocatechin have been the only commercially available
condensed tannin standards until very recently. Numerous flavone
and triterpenoid standards are commercially available, as are a
few flavanones. Acquiring compounds analogous to the condensed
tannins for use as internal standards is also difficult. Historically,
quercetin has been used as an internal standard when GC or liquid
chromatography has been used with purified condensed tannins.
However, because quercetin is present in many of the tissue types
that we have analyzed, it is inappropriate for whole samples. To
date, we have used three polyphenolic compounds as internal
standards: 2,4,4′-trihydroxybenzophenone, 2,4,2′4′-tetrahydroxy-
benzophenone, and hematoxylin (Figure 2). The primary qualities
desirable in an internal standard are similar recoveries relative to
the compounds of interest, clean chromatographic elutions, and
similar elution times to the compounds of interest. In this method,
the former is largely dependent on solubility in water as com-
pounds are partitioned from the aqueous phase into ethyl acetate.
In general, compounds with a higher number of hydroxyl groups
will have lower recoveries into ethyl acetate and vice versa. In
triplicate partition experiments involving the internal standards,
(20) Porter, L. J. In Natural Products of Woody Plants; Rowe, J. W., Ed.;
Springer: Berlin, 1989; pp 651-698.
(21) van der Westhuizen, J. H.; Steenkamp, J. A.; Ferreira, D. Tetrahedron 19 90 ,
46, 7849-7854.
5122 Analytical Chemistry, Vol. 72, No. 20, October 15, 2000

with the addition of sediment by a factor of 2. While many
mechanisms are possible, larger molecules have generally been
shown to be more reactive than smaller molecules toward mineral
surfaces in sorption experiments,22 and the trend in calculated
degree of polymerization is consistent. Tannin reactivity toward
the soil appears to be less than reactivity toward the sediment.
The addition of 1000 mg of soil gave yields corresponding to ∼400-
mg addition of sediment. The overall implication of this experiment
is that unreacted tannin may be detectable by GC in soils and
sediments at least as low as 0.03 wt %. Analysis of a soil sample
taken from within the dripline of the CLM camellia tree showed
the presence of tannin at 0.002 wt % as estimated by GC/ MS,
which shows greater sensitivity than GC.
To evaluate mechanisms for the matrix effects shown above,
POLL was hydrolyzed in the presence of combusted sediment
(to remove organic carbon), desalted combusted sediment, and
sodium chloride in amounts equivalent to that present in 1000
mg of sediment. Sodium chloride did not affect yields, while both
the combusted sediment additions resulted in no tannin. This
suggests that tannin is very mineral reactive and that organic
coatings might aid preservation. In addition to these matrix effects,
protein/ tannin interactions were also tested by the addition of 50
mg of bovine serum albumin to POLL prior to depolymerization.
Yields were not affected.
Our experience with mineral matrix effects was similar to that
of Schofield et al. in 1998.23 They were unable to detect tannin in
soils underlying willow leaf litter, despite evidence that tannin was
being leached from the leaves. Even when purified Sorghum tannin
was added to the soil at up to 0.2 wt %, they were unable to extract
any measurable tannin by several different techniques, including
sorptive techniques and direct depolymerization in the presence
of phloroglucinol.
Clearly, mineral matrix effects and the ability to quantify tannin
in soils and sediments are areas that need further study. There
are a number of factors to consider, the primary one being
whether tannin is sorbed by the minerals (and associated
organics) but remains intact or whether it is chemically altered.
If it is the former, then it may be possible in the future to find a
method for desorption. If it is the latter, the challenge will be to
determine how it is altered and how to quantify it. In either case,
if a means of extraction ultimately is not possible, then solid-phase
13C NMR or pursuing variations of stronger degradative tech-
niques such as pyrolysis or CuO oxidation may be necessary for
direct quantification. Mechanistically, sorption studies involving
purified tannin, soils, and sediments are in order.
ular tannin via this method was measured at 6% in both leaf
types, 7 and 10% with the Folin-Denis reagent, and ∼20% by 13
C
NMR.
Quantitative comparisons can also be made for bark tannins
of the same species analyzed by our method on whole samples18
and thioacidolysis (depolymerization in the presence of toluene-
R-thiol) on extracts.16 Barks from Pinus contorta and Pseudotsuga
menziesii were measured at the molecular level in both studies,
the latter method yielding 3.2 and 3.3% total tannin, respectively,
while our method gave 3.3 and 3.1%. The analysis by our method,
however, also included flavanones, which have been shown to be
present in gymnosperm bark tannin as terminal units27 and
constituted about half the tannin measured in these two samples.
While part of the difference is likely directly attributable to the
higher efficiency reported for thioacidolysis,14 preservation is also
likely a factor. Tannin in samples has been shown to become more
depolymerization resistant with age.14 The barks analyzed by
thioacidolysis were freshly collected26 whereas those by this
technique were >10 years old.
Compositional comparisons between the barks can also be
made. In the P. contorta and P. menziesii barks, the degree of
polymerization was measured at 4.7 and 3.7, respectively, by
thioacidolysis26 vs 4.4 and 2.5 for our technique.18 A commonly
measured parameter is the ratio of 2,3-cis (i.e., epicatechin and
epigallocatechin) monomers to 2,3-trans (i.e., catechin and gallo-
catechin) monomers. This ratio was determined to be 60:40 and
81:19, respectively, by thioacidolysis26 vs 60:40 and 64:36 by our
technique.18 Discrepancies in P. menziesii can be attributed to the
considerable variability found in inner and outer barks.26 Finally,
<20% of the condensed tannin in P. contorta bark as measured
by thioacidolysis was PD tannin, as compared to a previous
literature value of 69%.20 The measurement of 34% PD tannin by
our technique is intermediate.18
Compositional characteristics of tannin are commonly mea-
sured by 13C NMR on extracted and purified tannin. Results from
seed cones analyzed by our method18 can be compared to those
obtained by 13C NMR.28 13C NMR results from five Pinus sp. seed
cones indicated that 2,3-cis conformations made up 65-81% of
condensed tannin. Six seed cones were analyzed by our method,
with four species showing 62-90% 2,3-cis, and 37 and 38% in
Sequoia sempervirens and Tsuga heterophylla, respectively. The
only common species between the two sample sets was Pinus
ponderosa. In this species, 13C NMR indicated 74% 2,3-cis vs 62%
by our method. Degree of polymerization in the seed cones ranged
from 5.3 to 8.5 by 13C NMR vs 1.8-7.4 by our method (5.4 vs 2.4
for P. ponderosa). This latter difference is expected given that
monomers and smaller oligomers are eliminated in extraction and
purification procedures.
Comparisons to Other Methods. Analysis of a variety of
sources allows for direct comparisons to be made between this
and other tannin methods. Complete discussions can be found in
refs 18 and 24. In a mangrove leaf study, total tannin estimates
and measurements were made with 13C NMR and the Folin-Denis
reagent25 as well as with the method reported here.24 In green
and senescent yellow mangrove leaves, for instance, total molec-
On the whole, then, our method gives results that are
consistent with other established techniques for tannin, but
without extraction procedures and with much greater throughput.
Further Considerations. From the beginning of this project,
commercial availability of condensed tannin standards has been
(22) Thimsen, C. A.; Keil, R. G. Mar. Chem. 1 9 9 8 , 62, 65-76.
(23) Schofield, J. A.; Hagerman, A. E.; Harold, A. J. Chem. Ecol. 1 9 9 8 , 24, 1409-
1421.
(26) Matthews, S.; Mila, I.; Scalbert, A.; Donnelly, D. M. X. Phytochemistry 19 97 ,
45, 405-410.
(24) Hernes, P. J.; Benner, R.; Cowie, G. L.; Gon˜ i, M. A.; Bergamaschi, B. A.;
Hedges, J. I. Geochim. Cosmochim. Acta, in press.
(25) Benner, R.; Hatcher, P. G.; Hedges, J. I. Geochim. Cosmochim. Acta 1 9 9 0 ,
54, 2003-2013.
(27) Hergert, H. L. In Chemistry and Significance of Condensed Tannins;
Hemingway, R. W., Karchesy, J. J., Eds.; Plenum Press: New York, 1989;
pp 3-20.
(28) Eberhardt, T. L.; Young, R. A. J. Agric. Food Chem. 1 9 9 4 , 42, 1704-1708.
Analytical Chemistry, Vol. 72, No. 20, October 15, 2000 5123

a limiting factor in the overall method development. The catechin-
catechin dimer used to identify catechin-phloroglucinol adducts,
for instance, has only recently become available. Propelargonidin
standards do not exist. Authentic standards of phloroglucinol
adducts are still needed in order to improve quantification. In
addition, authentic standards of condensed tannin oligomers or
polymers are needed in order to evaluate depolymerization
efficiency. The catechin-catechin and catechin-epicatechin dimers
can certainly be useful in addressing this question. However, the
effect of chain length on depolymerization efficiency remains to
be seen, and epicatechin and catechin extender units are known
to have different kinetic rates of depolymerization.29
One measure of the effectiveness of a technique is how well it
performs under a given set of conditions, and the purpose of this
study has been to conduct such an evaluation. This technique
allows rapid, reproducible analyses of tannin and triterpenoids on
whole samples. An equally important measure, however, is the
potential of the method for improvement and adaptation. For
instance, identification and quantification of the carboxylic acids
will certainly increase the overall usefulness of the method.
Tetracosanoic acid as quantified by this method has already been
shown to be a potential indicator for barks.18 Possible modifications
to the method to increase sensitivity include on-column injection,
drying down the hydrolysate followed by direct extraction, or the
use of photoionization detectors tuned specifically to detect
phenolics.
Tannin in soils and sediments and the overall issue of mineral
matrix effects remain to be evaluated. However, this method has
already shown utility in the study of early diagenesis of mangrove
leaves24 and in characterizing nearly 120 source tissues at the
molecular level for tannin.18 As such, it may prove useful in many
geochemical studies involving terrestrial plant material as well as
in natural products.
ACKNOWLEDGMENT
The authors thank Caroline Preston for a valuable review and
perspective of this work, along with an anonymous reviewer.
Received for review November 11, 1999. Accepted July
19, 2000.
(29) Hemingway, R. W.; McGraw, G. W. J. Wood Chem. Technol. 1 9 8 3 , 3, 421-
435.
AC991301Y
5124 Analytical Chemistry, Vol. 72, No. 20, October 15, 2000