Discriminating between the six isomers of dicaffeoylquinic acid by LC-MSn

Article abstract of DOI:10.1021/jf050046h

The fragmentation behavior of all six dicaffeoylquinic acids (diCQA) has been investigated using LC-MS⁴. It is possible to discriminate between each of the isomers including those for which commercial standards are not available. For diCQA, the

Full text of DOI:10.1021/jf050046h

J. Agric. Food Chem. 2005, 53, 3821−3832 3821
Discriminating between the Six Isomers of Dicaffeoylquinic
Acid by LC-MSn
§
MICHAEL N. CLIFFORD,*,† SUSAN KNIGHT,† AND NIKOLAI KUHNERT
Centre for Nutrition and Food Safety and Department of Chemistry, School of Biomedical and Life
Sciences, University of Surrey, Guildford, Surrey GU2 7XH, United Kingdom
The fragmentation behavior of all six dicaffeoylquinic acids (diCQA) has been investigated using
LC-MS4. It is possible to discriminate between each of the isomers including those for which
commercial standards are not available. For diCQA, the ease of removal of the caffeoyl residue during
fragmentation is 1 ≈ 5 > 3 > 4. The distinctive fragmentation observed for the little-studied 1,4-
dicaffeoylquinic acid involves elimination of the C1 caffeoyl residue, repeated dehydrations leading
to the aromatization of the quinic acid moiety, and its decarboxylation. It is suggested that this process
is initiated by the C1 carboxyl protonating the C5 hydroxyl in the inverted chair conformer, followed
by its protonating the C1 caffeoyl residue in the favored chair conformation. The fragmentation of
1-caffeoylquinic acid is indistinguishable from that of the commercially available 5-caffeoylquinic acid,
but these two isomers can be distinguished easily by their facile chromatographic resolution on
reversed phase packings. The hierarchical key previously developed for characterizing chlorogenic
acids has been extended to accommodate 1-caffeoylquinic acid and the 1-acyl dicaffeoylquinic acids.
KEYWORDS: Asteraceae; caffeoylquinic acids; chlorogenic acids; coffee; Cynara; cynarin; dicaffeoyl-
quinic acids; LC-MSn
INTRODUCTION
Classically, chlorogenic acids (CGA) are a family of esters
formed between certain trans-cinnamic acids and (-)-quinic
acid [1L-1(OH),3,4/5-tetrahydroxycyclohexane carboxylic acid]
(1-3). Specimen structures are shown in Figure 1. CGA are
widely distributed in plants (2, 3); however, few commercial
standards are available (4), and precise identification of
individual CGA in complex mixtures is not easy. An LC-MSn
method developed for this purpose can discriminate between
the isomers of caffeoylquinic acid (CQA), p-coumaroylquinic
acid (pCoQA), feruloylquinic acid (FQA), dicaffeoylquinic acid
(diCQA), and caffeoyl-feruloylquinic acid (CFQA) that occur
in green robusta coffee beans (5). Key steps in this scheme are
as follows:
1. MSn for a monoacyl CGA corresponds to MSn+1 for a
related diacyl CGA.
2. Individual CGA subgroups can be distinguished by their
molecular ion [CGA - H+]-.
3. 4-Acyl CGA can be distinguished by an MS2 or MS3 base
peak at m/z 173.
4. 5-Acyl CGA can be distinguished by an MS2 or MS3 base
peak at m/z 191 accompanied by a weak (<5% of base peak)
[cinnamate - H+]- fragment.
Figure 1. Structure of caffeoylquinic and dicaffeoylquinic acids (IUPAC
numbering) (1).
5. 3-Acyl CGA can be distinguished by an MS2 or MS3
[cinnamate - H+]- base peak (FQA, pCoQA, and CFQA) or
by a combination of an MS2 or MS3 base peak at ∼m/z 191
accompanied by an intense (>50% base peak) [cinnamate -
H+]- fragment (CQA and diCQA).
* Author to whom correspondence should be addressed (fax 44 14 83
57 69 78; e-mail bss2mc@surrey.ac.uk or m.clifford@surrey.ac.uk).
† Centre for Nutrition and Food Safety.
§ Department of Chemistry.
10.1021/jf050046h CCC: $30.25
© 2005 American Chemical Society
Published on Web 04/14/2005

3822 J. Agric. Food Chem., Vol. 53, No. 10, 2005
Clifford et al.
Figure 2. Generation of 1,4-diCQA and 1-CQA: (a) relative yield of 1,3-
diCQA, 1,4-diCQA, and methyl caffeate produced by treating 1,5-diCQA
on ice with TMAH for various times; (b) relative yield of 1-CQA, 3-CQA,
4-CQA, and 1,4-diCQA produced by treating 1,3-diCQA with 2 M HCl at
100 °C.
Many plants, including coffee, produce CGA in which
esterification occurs at positions 3, 4, and 5 of the quinic acid
moiety, but not at position 1. However, 1-acyl CGA are found
in some Asteraceae, for example, artichoke, burdock, and arnica
(6-10), and it is desirable that this procedure be evaluated also
for 1-acyl CGA. This manuscript presents the results of this
study.
MATERIALS AND METHODS
CQA and DiCQA. A methanolic extract of green arabica coffee
beans was used as a convenient source of 3,4-diCQA, 3,5-diCQA, and
4,5-diCQA (VIII-X) as previously described (5). The coffee extract
was treated with Carrez reagents (1 mL of reagent A plus 1 mL of
reagent B) (11) to precipitate colloidal material, diluted to 100 mL
with 70% v/v aqueous methanol, and filtered through a Whatman no.
1 filter paper. The methanol was removed by evaporation with nitrogen
(N-Evap-111, Organomation Associates Inc., Berlin, MA) and the
aqueous extract stored at -12 °C until required, thawed at room
temperature, centrifuged (1360g, 10 min), and used directly for LC-
MS.
Cynarin (1,3-diCQA) (V) and 1,5-diCQA (VII) were obtained from
LGC Promochem (Hatfield, U.K.). Stock solutions (∼1 mg/mL) were
prepared in 10% methanol and diluted 100 times. 1,4-DiCQA (VI) was
prepared by interesterification with tetramethylammonium hydroxide
(TMAH) as previously described (5, 12, 13) using a 60 s treatment of
1,5-diCQA (VII) (∼10 µg/mL) on ice. The 1-acyl diCQA working
solutions were centrifuged (1360g, 10 min) and used directly for LC-
MS.
Figure 3. Specimen LC-MS chromatograms: (a) UV chromatogram (320
nm) and selected ion chromatogram (m/z 515) of the arabica coffee extract
spiked with TMAH-treated 1,5-diCQA (VII) on C18 packing; (b) UV
chromatogram (320 nm) and selected ion chromatogram (m/z 515) of
the arabica coffee extract spiked with TMAH-treated 1,5-diCQA (VII) on
phenylhexyl packing; (c) UV chromatogram (320 nm) and selected ion
chromatograms (m/z 353 and 515) of acid-treated 1,3-diCQA (V) on
phenylhexyl packing.

LC-MSn of Dicaffeoylquinic Acids
J. Agric. Food Chem., Vol. 53, No. 10, 2005 3823
Figure 4. Negative ion MSn spectra for 1-CQA and 3-CQA.
Table 1. Negative Ion MS3 Fragmentation Data for Caffeoylquinic Acids
MS2 ions
MS3 ions
parent
ion m/z
MS2 base
peak m/z
MS3 base
peak m/z
compd
N
m/z
inta
m/z
int
m/z
int
m/z
int
m/z
int
m/z
int
m/z
int
1-CQA
3-CQA
4-CQA
5-CQA
3
3
3
3
353.0
353.1
353.3
352.9
191.1
191.0
173.0
191.1
179.3
179.2
179.0
179.1
3
55
70
3
b.p.b
b.p.
191.1
b.p.
100
100
20
85.1
85.2
93.1
85.0
110.9
111.4
111.1
110.7
35
40
65
40
127.3
127.2
126.9
127.3
65
70
5
172.3
172.3
60
65
93.4
93.5
b.p.
40
40
100
50
135.1
135.1
135.1
10
15
3
100
30
172.3
45
93.1
a Int, intensity. b b.p., occurs as base peak.
Table 2. Negative Ion MS4 Fragmentation Data for Dicaffeoylquinic Acids
MS2 ions
parent
ion m/z
MS2 base
peak m/z
compd
N
m/z
inta
m/z
int
35
m/z
int
m/z
int
m/z
int
m/z
int
m/z
int
m/z
int
1,3-diCQA
1,4-diCQA
1,5-diCQA
3,4-diCQA
3,5-diCQA
4,5-diCQA
3
3
3
3
3
3
515.1
515.1
515.0
515.1
515.0
515.1
353.1
353.0
353.1
353.0
353.1
353.1
335.1
335.1
335.0
335.1
35
25
7
191.1
10
20
20
179.0
179.0
30
7
317.1
299.1
299.1
299.0
50
5
255.0
255.1
255.2
22
5
203.1
203.1
203.1
203.0
203.1
55
3
7
10
15
173.1
173.0
173.1
15
18
12
191.1
191.2
16
179.1
179.1
15
5
317.0
7
12
8
MS3 ions
MS4 ions
MS3 base
peak m/z
MS4 base
peak m/z
compd
N
m/z
int
m/z
int
m/z
int
5
m/z
int
m/z
int
m/z
int
60
m/z
int
m/z
int
1,3-diCQA
1,4-diCQA
1,5-diCQA
3,4-diCQA
3,5-diCQA
4,5-diCQA
3
3
3
3
3
3
191.1
173.2
191.0
173.1
191.1
173.1
179.1
179.1
179.0
179.1
179.2
179.2
60
70
7
91
53
80
b.p.b
191.1
b.p.
191.1
b.p.
100
27
100
53
135.2
135.1
10
20
85.1
93.0
85.1
93.1
85.2
93.1
127.2
127.1
127.0
127.1
127.0
127.0
78
5
60
3
95
3
172.3
110.9
111.1
110.8
111.1
110.9
111.4
35
45
55
50
60
38
93.1
b.p.
93.0
b.p.
93.1
b.p.
60
100
55
100
80
172.1
172.0
172.0
172.0
75
2
135.1
135.1
135.1
14
12
12
100
27
173.1
90
15
191.2
100
a Int, intensity. b b.p., occurs as base peak.
1-CQA (I) was prepared from 1,3-diCQA (V) by partial acid
hydrolysis. Stock solution (10 µL) was added to 2 M HCl (1 mL) in a
screw-cap test tube and heated in a boiling water bath for 2 h. The
products were cooled and centrifuged (1360g, 10 min) and used directly
for LC-MS.
and a PDA detector with a light-pipe flow cell (recording at 320, 280,
and 254 nm and scanning from 200 to 600 nm). This was interfaced
with an LCQ Deca XP Plus mass spectrometer fitted with an ESI source
(ThermoFinnigan) and operating in zoom scan mode for the accurate
determination of parent ion m/z and in data-dependent, MSn full scan
mode to obtain fragment ion m/z. Unless specified otherwise, the parent
ion or base peak ion was selected for further fragmentation. MS
LC-MSn. The LC equipment (ThermoFinnigan, San Jose, CA)
comprised a Surveyor MS pump, an autosampler with a 50 µL loop,

3824 J. Agric. Food Chem., Vol. 53, No. 10, 2005
Clifford et al.
Figure 5. Negative ion MS1 spectra for isomeric dicaffeoylquinic acids.
Figure 6. Negative ion MS2 spectra for isomeric dicaffeoylquinic acids.
operating conditions (negative ion) had been optimized using 5-CQA
(III) with a collision energy of 35%, an ionization voltage of 3.5 kV,
a capillary temperature of 350 °C, a sheath gas flow rate of 65 arbitrary
units, and an auxiliary gas flow rate of 10 arbitrary units.
ways, analyses were performed in which acetic acid was omitted from
the HPLC solvents.
RESULTS AND DISCUSSION
CGA separations were achieved on 150 × 3 mm columns containing
either Luna 5 µm phenylhexyl packing or Kromasil C18 5 µm
(Phenonemex, Macclesfield, U.K.). For routine separations solvent A
was water/acetonitrile/glacial acetic acid (980:20:5 v/v, pH 2.68);
solvent B was acetonitrile/glacial acetic acid (1000:5 v/v). Solvents
were delivered at a total flow rate of 300 µL min-1. The gradient profile
was 4% B to 33% B linearly in 90 min, a linear increase to 100% B
at 95 min, followed by 5 min isocratic, a return to 4% B at 105 min,
and 5 min isocratic to re-equilibrate. In experiments designed to
investigate ionization mechanisms and rationalize fragmentation path-
Chromatographic Data. All data for CGA presented in this
paper use the recommended IUPAC numbering system (1), and
specimen structures are presented in Figure 1. When necessary,
previously published data have been amended to ensure
consistency and avoid ambiguity.
The coffee extract gave a typical chromatogram, and 3,4-
diCQA, 3,5-diCQA, and 4,5-diCQA (VIII-X) were located by
their parent ion at m/z 515 and distinguished by their patterns

LC-MSn of Dicaffeoylquinic Acids
J. Agric. Food Chem., Vol. 53, No. 10, 2005 3825
Figure 7. Negative ion MS3 spectra for isomeric dicaffeoylquinic acids.
Figure 8. Negative ion MS4 spectra for isomeric dicaffeoylquinic acids.
of fragmentation as previously reported (5). These compounds
eluted from the C18 packing between 47 and 55 min and from
the phenylhexyl packing between 53 and 58 min. As previously
reported (14) 1,3-diCQA (V) eluted much earlier (∼27 min for
C18 and ∼31 min for phenylhexyl), whereas 1,5-diCQA (VII)
eluted at ∼48 min (C18) and ∼52 min (phenylhexyl).
Figure 2. These were identified as 1,3-diCQA (V) by com-
parison with an authentic standard and a second, less intense,
peak that on C18 eluted before 1,5-diCQA (VII) (∼45 min) but
after 1,5-diCQA (VII) on the phenylhexyl packing (∼55 min).
This new product had a parent ion at m/z 515 and MS3 and
MS4 base peaks at m/z 173 and 93, respectively. Such fragments
have previously been shown to be diagnostic for all 4-substituted
monoacyl and diacyl CGA so far examined (5).
Spiking of the TMAH-treated 1,5-diCQA (VII) into the green
coffee extract established that this peak was chromatographically
distinct from 3,4-diCQA (VIII) and 4,5-diCQA (X) on both
the C18 and the phenylhexyl packings, indicating that it must
be the expected 1,4-diCQA (VI) (Figure 3). The different
sequences of elution were not expected. In general, those CGA
with a greater number of free equatorial hydroxyl groups in the
Preliminary studies indicated that 1,5-diCQA (VII) was
rapidly and extensively converted to 1,3-diCQA (V) and 5-CQA
(III) by treatment with tetramethylammonium hydroxide
(TMAH), the substrate having almost disappeared after 1 min
at room temperature (22 °C). Because logically conversion of
1,5-diCQA (VII) to 1,3-diCQA (V) must proceed through 1,4-
diCQA (VI), the transesterification reaction conditions were
optimized by trial and error. The relative yields of the major
products as a function of time of treatment are presented in

3826 J. Agric. Food Chem., Vol. 53, No. 10, 2005
Clifford et al.

LC-MSn of Dicaffeoylquinic Acids
J. Agric. Food Chem., Vol. 53, No. 10, 2005 3827
Figure 9. Schemes showing the reactive intermediates.
quinic acid residue are more hydrophilic than those with a
greater number of free axial hydroxyl groups (13), but it is
known that on some packings 4-CQA (IV) precedes 5-CQA
(III), whereas on others the reverse occurs (4).
Preliminary studies indicated that treatment of 1,3-diCQA (V)
with hydrochloric acid (2 M, 2 h, 100 °C) caused extensive
degradation. Trial and error optimization of the acid treatment
(Figure 2b) with LC-MS of the products established that shorter
acid treatments produced 1,4-diCQA (VI) and three isomeric
CQA, each in a low yield.
1,5-DiCQA (VII) and 5-CQA (III) could not be detected
among the products. Two CQA isomers eluted from the
phenylhexyl packing at ∼12.8 and ∼14.5 min, respectively, well
in advance of 4-CQA (IV) at ∼22.7 min (Figure 3c). MS3
fragmentation (Figure 4) indicated that the peak at 14.5 min
was undistinguishable from 3-CQA (II) (5), and accordingly
the earlier eluting peak was assigned by a process of elimination
as the little studied 1-CQA (I).
Fragmentation of 1-CQA and the 1-Acyl DiCQA. The MSn
fragmentation data are summarized in Tables 1 and 2 for CQA
and diCQA, respectively. It is not possible to reliably distinguish
between 1-CQA and 5-CQA by their fragmentation. Fortunately,
5-CQA is readily available from commercial sources, and
1-CQA can be easily resolved from this, so, in practice,
discrimination is straightforward.
The fragmentations observed for 3,4-diCQA, 3,5-diCQA, and
4,5-diCQA (VIII-X) were identical to those previously reported
(5), but as discussed below reference is made in this paper to
some minor fragments not previously considered to be impor-
tant. All three 1-acyl diCQA (V-VII) gave the expected
molecular ion [diCQA - H+]- at m/z 515 and MS2 base peak
[diCQA-cinnamate - H+]- at m/z 353. Consistent with previous
studies (5) 1,3-diCQA (V) and 1,5-diCQA (VII) gave the
expected MS3 and MS4 base peaks at m/z 191 and 85,
respectively, whereas 1,4-diCQA (VI) yielded base peaks at m/z
173 and 93.
fragments at m/z 317 (∼35%) and m/z 255 (∼22%). The
fragment at m/z 203 is produced weakly (<15% base peak) by
all other isomers except 1,3-diCQA (V). The other three ions
are unique to 4-acyl diCQA but, in 3,4-diCQA (VIII) and 4,5-
diCQA (X), none exceed 15% of base peak. Because 3,4-diCQA
(VIII) has a comparatively intense MS2 fragment at m/z 335,
which is not seen in the MS2 spectrum of 4,5-diCQA (X) (in
Figure 10 of ref 5, this point is incorrectly stated but is correctly
illustrated in Figure 7), these factors collectively discriminate
between the three 4-acyl diCQA. The remaining three isomeric
diCQA (V, VII, and IX) can be distinguished by the relative
intensity of the MS2 fragment at m/z 335 (1,3-diCQA, ∼35%;
1,5-diCQA, ∼7%; and 3,5-diCQA, not detectable). The MS1,
MS2, MS3, and MS4 spectra are illustrated in Figures 5-8.
Previous studies (5) led to the conclusion that during diCQA
fragmentation the C5 caffeoyl is the most easily removed and
the C4 caffeoyl is the most stable, but the 1-acyl diCQA
molecules were not examined, and the relative stability of the
C1 caffeoyl could not be estimated. If the C1 caffeoyl were the
most stable, then the MS2 and MS3 fragmentations for the three
1-acyl diCQA (V-VII) would be identical because the MS2
base peak would be [1-CQA - H+]- in each case. From Table
1 and the discussion above, it is clear that this is not the case.
Comparison of MS2 and MS3 fragmentation data (Table 1) for
the CQA (I-IV) with the MS3 and MS4 fragmentation for the
1-acyl diCQA (Table 2) indicates perfect matches for 1,3-
diCQA (V) with 3-CQA (II), for 1,4-diCQA (VI) with 4-CQA
(IV), and for 1,5-diCQA (VII) with both 1-CQA (I) and 5-CQA
(III). This demonstrates clearly that the caffeoyl moiety at C1
is more easily removed than the caffeoyl residues at C3 or C4,
but because 1-CQA (I) and 5-CQA (III) fragment identically,
it is not possible to determine whether the MS2 base peak
produced from 1,5-diCQA (VII) is [1-CQA - H+]- or [5-CQA
- H+]-, and thus one cannot judge on this basis which caffeoyl
residue is more easily removed.
Mechanisms of Fragmentation. We have reconsidered the
mechanisms of fragmentation and no longer favor the radical
anion pathway originally proposed (5). We now suggest that in
order to rationalize the observed fragmentations, two alternative
From Table 2, 1,4-diCQA (VI) is easily distinguished by its
unique MS2 fragments including comparatively intense (>50%
base peak) ions at m/z 299 and 203 supported by less intense

3828 J. Agric. Food Chem., Vol. 53, No. 10, 2005
Clifford et al.
Figure 10. 1H 1H COSY NMR spectrum (500 MHz) of 1,3-diCQA (V) in D2O.
−
pathways must be considered. Which pathway operates depends
crucially upon the site of the negative charge within the
molecular ion. Although the initial ionization will occur at the
most acidic moiety within the molecule (the COOH in the case
of all quinic acid derivatives), it may subsequently migrate. In
either case it leads to the formation of an anion [M - H+]- at
m/z 515. Because a priori we cannot eliminate either possibility,
we therefore present both pathways and discuss in detail their
individual merits.
The alternative scenario envisages fragmentation driven by
an anion located other than on the quinic carboxyl. Although
an initial ionization at this other site cannot be excluded
unequivocally, the alternative anion is more likely to form after
migration of the negative charge from the carboxyl by inter- or
intramolecular proton transfer. As a representative example
Figure 9, Scheme A, shows the negative charge located on a
particular phenolic oxygen F, but other sites for the negative
charge, such as G, cannot be excluded. Such proton transfers
to non-carboxylic sites have been postulated (15) for serine,
threonine, or tyrosine residues during ESI mass spectrometry
of proteins. The C1 carboxyl and the C5 caffeoyl again have
the 1,3-syn-diaxial arrangement (Figure 9, Scheme A, pathway
B), and the comparatively easy loss of the C5 caffeoyl can be
rationalized (5) by transfer of the proton from the C1 carboxyl,
while in an inverted chair, to the acyl oxygen of the C5
substituent as shown in F and G. Deacylation facilitated by
protonation of the C5 acyl substituent would give neutral caffeic
acid H and, after proton transfer, anionic fragment I with m/z
335.
When the negative charge is located at the carboxyl in A,
we suggest that the comparatively easy loss of the C5 caffeoyl
residue can be rationalized by assuming an acyl transfer
mechanism (Figure 9, Scheme A, pathway A). In a 1,3-syn-
diaxial chair conformation the COO- acts as a nucleophile and
attacks the C5 acyl carbonyl to form a bicyclic tetrahedral
intermediate B. Loss of a proton from the p-OH on the 5-acyl
moiety (most likely via an intramolecular proton shift because
no dianion fragments could be observed in any mass spectra)
leads to a breakdown of the tetrahedral intermediate to give C.
Further fragmentation leads to the formation of the deacylated
quinic acid fragment ion D and most likely a neutral quinone
ketene E.
The comparatively easy loss of the C1 acyl group could be
explained by an analogous mechanism (Figure 9, Scheme B,

LC-MSn of Dicaffeoylquinic Acids
J. Agric. Food Chem., Vol. 53, No. 10, 2005 3829
Figure 11. 1H 13C HMBC NMR (500 MHz) of 1,3-diCQA (V) in DMSO-d6 expanded to show aromatic, olefinic, and phenolic protons.
−
pathway A). Again, after initial formation of COO- to give J,
nucleophilic attack at the C1 acyl carbonyl leads to the formation
of a spirocyclic tetrahedral intermediate K, which after break-
down of the tetrahedral carbon gives L. Loss of a proton (again
presumably via an intramolecular proton shift) gives a de-
acylated fragment ion M and a neutral quinone ketene E. Both
modes of attack (7-exo-trig for attack at the C5 acyl moiety
and 5-exo-trig for attack at the C1 acyl moiety) are allowed by
Baldwin rules. Because we never observed a fragment ion at
m/z 161, corresponding to [E - H]-, it can be inferred that the
negative charge in C and L is not transferred to the ketene E.
In the 1-acyl diCQA (V-VII) we suggest again as an
alternative pathway that the C1 carboxyl can, after proton
transfer to give N, protonate the C1 caffeoyl group (Figure 9,
Scheme B, pathway B). Because this occurs in the preferred
carboxy-equatorial conformation of the quinic acid moiety, the
C1 caffeoyl might be more rapidly removed than the C5 caffeoyl
that requires the thermodynamically less favored carboxy-axial
conformation. Deacylation from N is again facilitated by
protonation of the C1 acyl moiety and gives fragment ion O
and neutral caffeic acid H.
-
Figure 12. Infrared spectrum (4000 to 800 cm 1) of 1,3-diCQA (V) in
Nujol.
in the chemical shifts for the two acyl moieties. In each of the
NMR spectra the resonances for the two caffeoyl residues appear
at distinct chemical shifts. 1H-1H COSY analysis (Figure 10)
and H-13C HMBC spectra (Figure 11) clearly indicate that
1
This hypothesis is supported by a close inspection of NMR
the olefinic signals of one caffeoyl residue are shifted signifi-
cantly downfield compared with the signals for the second
caffeoyl residue. This can be interpreted in terms of hydrogen
bonding of the C1 carboxyl to one of the caffeoyl residues (17,
1
and IR spectral data for 1,3-diCQA (V) (16). In the H NMR
and 13C NMR spectra of 1,3-diCQA (V) in a variety of solvents,
such as D2O, DMSO-d6 and MeOH-d4, differences are observed

3830 J. Agric. Food Chem., Vol. 53, No. 10, 2005
Clifford et al.
Figure 13. Structures of the dicaffeoylquinic acid derived fragments.
1
18). Unfortunately, the low resolution of the H-13C HMBC
four broad bands around 3390 cm-1 and a sharp band at 3596
cm-1 are observed. Although hydrogen bonding is usually
associated with broadened bands in the IR spectrum, there are
a series of literature examples in which strong intramolecular
hydrogen bonds are manifested by sharp bands at higher
wavenumbers (19-21). It is worth noting that this sharp band
is absent in the IR spectrum of 5-CQA (III) (22). The carbonyl
region in the IR spectrum of 1,3-diCQA (V) strongly supports
the hypothesis of a C1 carboxyl-ester oxygen hydrogen bond.
Three CdO absorptions (Figure 12) are observed in the
carbonyl region at 1715, 1683, and 1638 cm-1. The absorption
at 1715 cm-1 can be assigned to the ester CdO of the C3
caffeoyl by comparison with the CdO absorption of trans-
methylcinnamate at 1722 cm-1 (23). The absorption at 1638
cm-1 can be assigned to the CdO absorption of the C1 carboxyl
by comparison with the CdO absorption in 5-CQA (III) at 1640
spectrum does not allow unambiguous identification of the two
CdO resonances connected to the olefinic moieties. In the H
1
NMR spectrum in DMSO-d6, the resonance for the carboxyl
COOH proton is observed at 13.8 ppm, considerably downfield
compared with this signal in quinic acid (Q) and 3-CQA (II),
again indicating hydrogen bonding (17, 18). Molecular modeling
at the MM-2 level indicates that for 1-CQA (I) there are two
minimized energy structures involving hydrogen bonds between
the equatorial C1 carboxyl (COOH) and the axial C1 caffeoyl
residue that differ by only 0.5 kJ‚mol-1. In the first structure,
the hydrogen bonds to the CdO carbonyl oxygen and in the
second to the O-CdO ester oxygen.
Furthermore, the IR spectrum (Figure 12) of 1,3-diCQA (V)
is consistent with a hydrogen bond between the equatorial C1
carboxyl and the axial C1 caffeoyl residue. In the OH region,

LC-MSn of Dicaffeoylquinic Acids
J. Agric. Food Chem., Vol. 53, No. 10, 2005 3831
Figure 14. Hierarchical key for the identification by LC-MSn of caffeoylquinic and dicaffeoylquinic acids including those substituted at position 1.
cm-1. Therefore, by a process of elimination, the third CdO
absorption at 1683 cm-1 has to be assigned to the CdO
absorption of the C1 caffeoyl residue. From the observed
wavenumber it clearly follows by comparison with trans-
methylcinnamate and 5-CQA that the absorption is shifted
toward lower wavenumbers, indicative of a weakening of the
CdO bond and hence hydrogen bonding between the equatorial
C1 carboxyl and the axial C1 caffeoyl residue.
Additional information on the two alternative fragmentation
mechanisms for loss of a caffeoyl moiety from either C1 or C5
has been obtained from an experiment in which the pH value
was varied. For a proton transfer process the intensities of
fragment ions observed should increase at a lower pH. In
contrast, for a COOH deprotonation ionization mechanism the
intensities of fragment ions observed should increase at a higher
pH of the ESI solution. By omitting the acetic acid cosolvent
from the LC-MS experiment, the chromatographic resolution
of the CQAs was affected; however, no change in the relative
intensities of the fragment ions was observed within experi-
mental error.
Although the fragmentations observed for 1-CQA (I) and
5-CQA (III) are indistinguishable one from the other, they are
quite different from the fragmentations previously observed for
3-CQA (II) and 4-CQA (IV) (5). This is in keeping with the
fragmentation of 3-CQA (II) and 4-CQA (IV) occurring via
mechanisms quite different from nucleophilic attack by COO-
or protonation by the quinic acid carboxyl operative for the
caffeoyl residues at C1 and C5.
Fragmentations of 3-CQA and related diCQA can be ex-
plained by arguments similar to those above (Figure 9, Scheme
C, pathway A). The only 1,3-syn-diaxial arrangement available
to the caffeoyl substituent in 3-CQA (II) involves the C1
hydroxyl (Figure 9, Scheme C, pathway A), and the high pK
of the hydroxyl (pK >15) relative to the carboxyl (pK ≈ 3.5)
explains why 3,5-diCQA (IX) loses the residue at C5 prior to
the residue at C3. Starting from the [M - H]- ion P, protonation
of the C3 acyl group by the C1 hydroxyl group would give
fragment ion R and neutral caffeic acid H. Alternatively, after
intramolecular proton transfer, the negative charge could be
located at the C1 oxygen in S, which after nucleophilic attack
on the C3 acyl carbonyl carbon would give tetrahedral inter-
mediate T. Breakdown of this intermediate would lead via U
to the deacylated anion V and quinone ketene E. It is worth
pointing out that the fragmentation mechanisms suggested here
constitute more detailed extensions of those reported previously
(5).
In summary, the ease of loss for caffeoyl moieties esterified
to quinic acid is determined by the relative stereochemical
relationships between the caffeoyl substituents and the other
functionalities, in particular the C1 carboxyl and C1 hydroxyl.
The comparatively facile loss of a caffeoyl moiety from C1 or
C5 of the diCQAs can be rationalized in terms of an acyl transfer
mechanism initiated by a nucleophilic carboxylate COO- or
alternatively by a protonation mechanism involving the acidic
COOH at C1. The comparatively more difficult loss of a caffeoyl
moiety from C3 can be rationalized in terms of protonation of
the C3 caffeoyl substituent by the C1 hydroxyl or alternatively
by an acyl transfer mechanism involving the nucleophilic C1
alcoholate. Loss of water is preferred for a caffeoyl substituent
at C4 as discussed previously (5). All proposed mechanisms
are in agreement with the observed mass spectra and constitute
a very useful tool for the prediction of fragmentation patterns
in structurally related molecules. Further experiments are
required to distinguish unambiguously between alternative
mechanisms.
Whatever the pathway(s), the formation of “dehydrated”
fragments [Figure 13 m/z 335 (Q2) and m/z 173 (Q1)] is
characteristic of 4-acyl CGA (5), and the fragments at m/z 317

3832 J. Agric. Food Chem., Vol. 53, No. 10, 2005
Clifford et al.
(6) Slanina, J.; Paulova´, H.; Humpa, O.; Bochora´kova´, H.; Tabo-
ra´ska´, E. 1,5-dicaffeoylquinic acid, and antioxidant component
of Cynara cardunculus leaves. Scr. Med. 1999, 72, 9-18.
(7) Maruta, Y.; Kawabata, J.; Niki, R. Antioxidative caffeoylquinic
acid derivatives in the roots of burdock (Arctium lappa L.). J.
Agric. Food Chem. 1995, 43, 2592-2595.
(8) Ben Hod, G.; Basnizki, Y.; Zohary, D.; Mayer, A. M. Cynarin
and chlorogenic acid content in germinating seeds of globe
artichoke (Cynara scolymus L.). J. Genet. Breed. 1992, 46, 63-
68.
(9) Merfort, I. Caffeoylquinic acids from flowers of Arnica montana
and Arnica chamossonis. Phytochemistry 1992, 31, 2111-
2113.
(10) Horman, I.; Badoud, R.; Ammann, W. Food-related applications
of one- and two-dimensional high-resolution proton nuclear
magnetic resonance: structure and conformation of cynarin. J.
Agric. Food Chem. 1984, 32, 538-540.
(11) Egan, H.; Kirk, R. S.; Sawyer, R. Pearson’s Chemical Analysis
of Foods; Churchill Livingstone: London, U.K., 1981.
(12) Clifford, M. N.; Kellard, B.; Birch, G. G. Characterisation of
caffeoylferuloylquinic acids by simultaneous isomerisation and
transesterification with tetramethylammonmium hydroxide. Food
Chem. 1989, 34, 81-88.
(13) Clifford, M. N.; Kellard, B.; Birch, G. G. Characterisation of
chlorogenic acids by simultaneous isomerisation and trans-
esterification with tetramethylammonium hydroxide. Food Chem.
1989, 33, 115-123.
(Figure 13, Q3) and m/z 299 seen clearly during fragmentation
of 1,4-diCQA (V) continue this sequence, leading to full
aromatization of the quinic acid residue and a caffeoylbenzoic
acid structure (Figure 13, Q4). Such dehydration has been
postulated (24) during pyrolysis of 5-CQA (III). The fragment
at m/z 255 probably arises by subsequent loss of carbon dioxide
(Figure 13, Q5) from the caffeoylbenzoic acid. The fragment
at m/z 203 was trapped and further fragmented, yielding
progressively m/z 175, 147, and 119 as base peaks, the latter
accompanied by m/z 129 (∼45%).
The peak at m/z 203 could result from loss of C4H4 from Q5.
This fragmentation is highly unusual and has no literature
precedent to our knowledge. It might involve a rearrangement
followed by a retro-Diels-Alder process. The fragments at m/z
175, 147, and 119 could form from the sequential consequential
loss of three CO units. Loss of CO from esters and phenols is
well precedented in EI mass spectrometry. The peak at m/z 129
should occur due to loss of water from m/z 147.
Dehydration accompanying elimination of a moiety at C4 is
thought to involve a conformer that allows 1,2-acyl participation
by the trans-vicinal acyl moiety on C4, thus facilitating the
formation of a bicyclic oxonium radical anion by loss of OH
from C5 assisted by protonation in the inverted chair conforma-
tion and subsequent loss of the caffeoyl at C4 or C1. For
chemical arguments we favor elimination of the caffeoyl at C1.
Loss of the C4 caffeoyl would lead to an enolether-enone pair
of tautomers, which are not amenable to further dehydration
and aromatization. Furthermore, retention of the caffeoyl at C1
would require caffeoyl migration prior to aromatization. In
contrast, loss of the C1 caffeoyl would, after subsequent
dehydrations involving loss of the C3 OH from fragment (Q3),
produce the benzoic acid (Q4) fragment ion. To obtain a final
proof of this hypothesis would require independent synthesis
and investigations into the fragmentation of Q2-Q5, and is
beyond the scope of the present investigation.
Revised Hierarchical Key for DiCQA. Figure 14 incorpo-
rates revisions to the hierarchical key (5) developed for CGA
not substituted at position 1. It is clear that all six of the diCQA
can be distinguished by LC-MS3 with further confirmation
furnished at MS4. 1-CQA and 5-CQA cannot be distinguished
by MS fragmentation, but the greater hydrophobicity of 5-CQA
and its commercial availability ensure that the two can be
distinguished by retention time on reversed phase packings.
(14) Clifford, M. N. Coffee bean dicaffeoylquinic acids. Phytochem-
istry 1986, 25, 1767-1769.
(15) Loo, J. A.; Ogorzalek-Loo, R. R.; Light, K. J.; Edmonds, C. G.;
Smith, R. D. Multiply charged negative ions by electrospray
ionization of polypeptides and proteins. Anal. Chem. 1992, 64,
81-88.
(16) Slanina, J.; Ta´borska´, E.; Bochora´va´, H.; Slaninova´, I.; Humpa,
O.; Robinson, W. E.; Schram, K. H. New and facile method of
preparation of the anti-HIV-1 agent, 1,3-dicaffeoylquinic acid.
Tetrahedron Lett. 2001, 42, 3383-3385.
(17) Emsley, J. W.; Feeney, J.; Sutcliffe, L. H. 9. The effects of
chemical equilibria and molecular connformational motion on
NMR spectra. In High-Resolution Nuclear Magnetic Resolution
Spectroscopy; Pergamon Press: Oxford, U.K., 1965; pp 481-
588.
(18) Kuhnert, N.; Le Gresley, A. The synthesis of tetra-acrylamido-
calix[4]arene capsules. Chem. Commun. (Cambridge) 2003, 19,
2426-2427.
(19) Eames, J.; Kuhnert, N.; Warren, S. The scope and limitations of
the [1,2]-alkylsulfonyl (SMe, SEt, and SCH2Ph) and sulfanyl
(SH) migration in the stereospecific synthesis of substituted
tetrahydrofurans. J. Chem. Soc., Perkin Trans. 1 2001, 1504-
1511.
(20) Eames, J.; Kuhnert, N.; Warren, S. The scope and limitation of
the [1,4]-S-benzyl participation and debenzylation in the stereo-
chemically controlled synthesis of substituted thiolanes. J. Chem.
Soc., Perkin Trans. 1 2001, 138-143.
(21) Davies, M. M. Infrared Spectroscopy and Molecular Structure:
An Outline of the Principle; Elsevier: Amsterdam, The Neth-
erlands, 1965.
(22) Sefkow, M.; Kelling, A.; Schilde, U. First efficient syntheses of
1-, 4-, and 5-caffeoylquinic acid. Eur. J. Org. Chem. 2001,
2735-2742.
ACKNOWLEDGMENT
Technical assistance from H. Roozendaal is gratefully acknowl-
edged.
LITERATURE CITED
(1) IUPAC. Nomenclature of cyclitols. Biochem. J. 1976, 153, 23-
31.
(2) Clifford, M. N. Chlorogenic acids and other cinnamatessnature,
occurrence, dietary burden, absorption and metabolism. J. Sci.
Food Agric. 2000, 80, 1033-1042.
(3) Clifford, M. N. Chlorogenic acids and other cinnamatessnature,
occurrence and dietary burden. J. Sci. Food Agric. 1999, 79,
362-372.
(4) Clifford, M. N. The analysis and characterization of chlorogenic
acids and other cinnamates. In Methods in Polyphenol Analysis;
Santos-Buelga, C., Williamson, G., Eds.; Royal Society of
Chemistry: Cambridge, U.K., 2003; Chapter 14, pp 314-337.
(5) Clifford, M. N.; Johnston, K. L.; Knight, S.; Kuhnert, N. A
hierarchical scheme for LC-MSn identification of chlorogenic
acids. J. Agric. Food Chem. 2003, 51, 2900-2911.
(23) The Aldrich Library of Infrared Spectra; Pouchert, C. J., Ed.;
Aldrich Chemical Co.: Milwaukee, WI, 1981.
(24) Lorant, B. Die thermische Zersetzung lebensmittelchemisch
interessanter Inhaltstoffe von Kaffee und Kakao. Nahrung 1968,
12, 351-356.
Received for review January 10, 2005. Revised manuscript received
March 14, 2005. Accepted March 23, 2005.
JF050046H