Regioisomeric characterization of triacylglycerols using silver-ion HPLC/MS and randomization synthesis of standards

Article abstract of DOI:10.1021/ac900150j

Silver-ion normal-phase high-performance liquid chromatography (HPLC) provides a superior separation selectivity for lipids differing in the number and position of double bonds in fatty acid chains including the resolution of triacylglycerol (TG) regioiso

Full text of DOI:10.1021/ac900150j

Anal. Chem. 2009, 81, 3903–3910
Regioisomeric Characterization of Triacylglycerols
Using Silver-Ion HPLC/MS and Randomization
Synthesis of Standards
Miroslav Lı´sa, Hana Velı´nska´ , and Michal Holcˇ apek*
ˇ
Department of Analytical Chemistry, Faculty of Chemical Technology, University of Pardubice, Na´m. Cs. Legi´ı 565,
53210 Pardubice, Czech Republic
Silver-ion normal-phase high-performance liquid chro-
matography (HPLC) provides a superior separation se-
lectivity for lipids differing in the number and position of
double bonds in fatty acid chains including the resolution
of triacylglycerol (TG) regioisomers under optimized
conditions. Our silver-ion HPLC method is based on the
coupling of three columns in the total length of 75 cm and
a new mobile phase gradient consisting of hexane-aceto-
nitrile-2-propanol which provides better resolution and
also reproducibility in comparison to previously used
mobile phases. In our work, the chemical interesterifica-
tion (randomization) of single-acid TG standards is used
for the generation of regioisomeric series of TGs, because
it provides a random distribution of fatty acids in TGs at
well-defined concentration ratios. The baseline separation
of regioisomeric TG pairs containing up to three double
bonds and the partial separation of TG regioisomers with
four to seven double bonds are reported for the first time.
Our silver-ion high-performance liquid chromatography/
mass spectrometry (HPLC/MS) method is applied for the
regioisomeric characterization of complex samples of
plant oils and animal fat, where the results clearly
demonstrate different preference of sn-2 occupation in
plants (mainly unsaturated fatty acids) versus animal fat
(mainly saturated fatty acids).
human body depending on sn positions. Concerning the overall
characterization of TGs in complex natural mixtures, the best
results are obtained with nonaqueous reversed-phase (NARP)
high-performance liquid chromatography (HPLC) coupled to mass
spectrometry (MS), where the highest number of identified TGs
has been reported for the coupling of chromatographic columns
in the total length of 45 cm and the mobile phase gradient
consisting of acetonitrile-2-propanol.5-7 The retention depends
on the equivalent carbon number (ECN) defined as the total CN
in all acyl chains minus 2 times the number of DBs, but the
optimized NARP systems can resolve most TGs even within one
ECN group with the exception of positional isomers. The partial
separation of regioisomers in NARP mode is feasible only with
multiple column coupling and very long retention times in the
range of 100-200 min,8,9 which is not practical for routine analysis.
The main possibilities of regiospecific analysis of TGs are the
following: (1) silver-ion HPLC,10-13 (2) MS,5-7 (3) enzymatic
hydrolysis (e.g., pancreatic lipase, phospholipase A2)
14 followed
by some analytical technique (e.g., silver-ion HPLC), and (4)
derivatization15 followed by chiral HPLC (complicated, labori-
ous, not suitable for complex mixtures). Silver-ion normal-phase
HPLC is a powerful technique for the separation of lipids
differing in the number and position of DBs. The principle of
this method is based on the capability of unsaturated organic
compounds to create weak reversible complexes with transition
metals, such as a silver ion. Silver ions are integrated into the
silica stationary phase (preferably on the basis of ion-ex-
change mechanism) interacting with π electrons of DBs during
the sample elution throughout the chromatographic column.
The retention of each sample compound primarily depends
Triacylglycerols (TGs) are the main constituents of plant oils1-3
and animal fats4 characterized by the total carbon number (CN),
the type and stereospecific position of fatty acids, and the number,
position, and configuration of double bonds (DBs) in acyl chains.
Fatty acids are digested in the human or animal organism with
the assistance of the stereospecific lipases, where fatty acids from
sn-1 and sn-3 position are cleaved first yielding 2-monoacylglyc-
erols. The stereospecific analysis of TGs is a long-standing and
challenging problem in the lipid analysis, but it is of high
importance due to the different availability of fatty acids for the
(5) L´ısa, M.; Holcˇapek, M. J. Chromatogr., A 2008, 1198-1199, 115–130
(6) Holcˇapek, M.; Jandera, P.; Fischer, L.; Prokesˇ, B. J. Chromatogr., A 1999,
858, 13–31
(7) Holcˇapek, M.; L´ısa, M.; Jandera, P.; Kaba´tova´, N. J. Sep. Sci. 2005, 28,
1315–1333
(8) Momchilova, S.; Itabashi, Y.; Nikolova-Damyanova, B.; Kuksis, A. J. Sep.
Sci. 2006, 29, 2578–2583
(9) Momchilova, S.; Tsuji, K.; Itabashi, Y.; Nikolova-Damyanova, B.; Kuksis,
A. J. Sep. Sci. 2004, 27, 1033–1036
(10) Christie, W. W. J. Chromatogr. 1988, 454, 273–284
(11) Schuyl, P. J. W.; De Joode, T.; Vasconcellos, M. A.; Duchateau, G. S. M. J.
E. J. Chromatogr., A 1998, 810, 53–61
(12) Adlof, R. O. J. High Resolut. Chromatogr. 1995, 18, 105–107
(13) Adlof, R. O.; List, G. J. Chromatogr. A 2004, 1046, 109–113
(14) Janssen, H. G.; Hrncˇirˇ´ık, K.; Szo´ra´di, A.; Leijten, M. J. Chromatogr., A 2006,
1112, 141–147
(15) Kuksis, A.; Itabashi, Y. Methods 2005, 36, 172–185
.
.
.
.
* Corresponding author.
(1) FEDIOL, The EU Oil and Proteinmeal Industry, Brussels, Belguim. http://
www.fediol.be (accessed Dec 2008).
(2) Scheeder, M. In Lipids from Land Animals in Modifying Lipids for Use in
Food; Gunstone, F. D., Ed.; Woodhead Publishing: Cambridge, 2006.
(3) Leray, C. Cyberlipid Center, Paris. http://www.cyberlipid.org (accessed
Dec 2008).
(4) Food and Agricultural Organization of the United Nations (FAO), Rome.
http://www.fao.org (accessed Dec 2008).
.
.
.
.
.
.
.
10.1021/ac900150j CCC: $40.75  2009 American Chemical Society
Published on Web 04/14/2009
Analytical Chemistry, Vol. 81, No. 10, May 15, 2009 3903

on the number of DBs, but it is also affected by the steric
availability of DBs for the interaction with silver ions. The
retention increases with increasing number of DBs with the
secondary separation according to the position and geometry
of DBs. Off-line or online two-dimensional (2D) HPLC of
relatively complementary separation modes in NARP and silver-
ion HPLC is a highly promising method for the analysis of
The chemical interesterification (so-called randomization) has
been used in many industrial applications in fat and oil process-
ing.28 It changes physicochemical properties of natural oils with
the assistance of fatty acids already present in TGs in a given oil
or fat. The degree of change is based on the reaction temperature,
reaction time, and catalysts used. The most common catalyst in
this process is sodium methoxide, but it is possible to use bases,
acids, and some metal ions as well. Sodium methoxide has the
highest reactivity, but on the other hand, it is very sensitive to
any trace of water, which stops the reaction due to the hydrolysis
of sodium methoxide. In the first reaction step, the bonds between
glycerol and fatty acids are cleaved yielding a mixture of diacylg-
lycerols (DGs), monoacylglycerols (MGs), and fatty acids. These
species subsequently undergo the interesterification reactions
providing a random distribution of fatty acids in newly formed
TGs.
The main goal of our work is the development of a silver-ion
HPLC/MS method applicable for the separation and quantitation
of regioisomeric ratios of TGs in complex mixtures. TG regioi-
someric pairs are separated by the optimized silver-ion HPLC
method, identified based on characteristic differences in their
APCI mass spectra, and quantified according to the ratio of
chromatographic peak areas. The randomization procedure plays
an important role in generating the series of regioisomeric
standards. Applications to complex natural samples of plant oils
and animal fats containing different TG regioisomers are
demonstrated.
,17
TGs.16
The different ratios of [M + H - RiCOOH]+ fragment ions
in atmospheric pressure chemical ionization (APCI) mass
spectra of positional isomers (regioisomers) of TGs was first
reported in 199618 and later on applied for HPLC/APCI-MS
characterization of prevailing fatty acids in sn-2 position in plant
oils.19 All MS approaches are based on the fact that the neutral
loss of fatty acid from the sn-2 position yields the fragment ion
with a lower relative abundance compared to cleavages from
the side sn-1/3 positions. It is often applied for the assignment
of prevailing fatty acids in the sn-2 position, but for the
quantitative determination of sn-2 occupation the calibration
curves for mixtures of both regioisomers have to be measur-
ed20-23 using the same instrument and ionization technique. APCI
is the most frequently used ionization technique for TG analysis
due to their nonpolar character, but electrospray ionization (ESI)
can be applied as well due to the formation of ammonium
adducts.24-26
If the TG has different fatty acids in sn-1 and sn-3 positions,
then the carbon atom in the sn-2 position becomes a chiral center.
Common analytical techniques working in a nonchiral environ-
ment cannot differentiate between sn-1 and sn-3 enantiomers;
therefore, they are generally treated as equivalent due to the lack
of suitable analytical techniques for their analysis.3,5,19,21 No chiral
separation of intact TGs have been reported so far with the
exception of synthetic TGs containing very different fatty acids
C8:0 versus C22:5 or C22:6,27 which is not a combination occurring
in nature. Due to the absence of any official recommendation for
the designation of sn-1/3 isomers, we list them in the order of
decreasing masses in accordance with our previously proposed
rule,20 e.g., OLP but not PLO. In case of isobaric fatty acids in
sn-1/3 positions, the more common fatty acid is listed first, e.g.,
LnOγLn but not γLnOLn.
EXPERIMENTAL SECTION
Materials. Acetonitrile, 2-propanol, methanol, ethanol, propi-
onitrile, ethylacetate, hexane (solvents are HPLC gradient grade),
and sodium methoxide were purchased from Sigma-Aldrich (St.
Louis, MO). Standards of tripalmitin (PPP, C16:0), triolein (OOO,
∆9-C18:1), trilinolein (LLL, ∆9,12-C18:2), and trilinolenin (LnLnLn,
∆9,12,15-C18:3) were purchased from Nu-ChekPrep (Elysian,
MN). Palm and olive oils were purchased from Augustus Oil
Limited (Hampshire, England).
Sample Preparation. An amount of 10-15 g of the sample
(sunflower or blackcurrant seeds, fat tissue from pig) was
weighed, and then seeds were carefully crushed in a mortar to
fine particles, whereas fat tissue was crushed in a homogenizer.
Then 15 mL of hexane was added, and this mixture was stirred
occasionally for 15 min. The solid particles were filtered out using
a course filter paper, and the extract was filtered again using a
fine filter (0.45 µm). From the filtered extract, hexane was
evaporated using a mild stream of nitrogen to yield pure plant oil
or animal fat.
Randomization. Amounts of 50 mg of each TG standard and
100 mg of sodium methoxide were weighed into a dry boiling
flask with the addition of 2 mL of hexane dried with molecular
sieves. The mixture was heated for 30 min in a water bath under
the reflux condenser. The reaction temperature was kept constant
at 75 °C. Then, the mixture was extracted with water and three
times with 1 mL of methanol to remove sodium methoxide. The
hexane phase containing randomized analyte was injected into
the silver-ion HPLC system.
(16) Dugo, P.; Favoino, O.; Tranchida, P. Q.; Dugo, G.; Mondello, L. J. Chro-
matogr., A 2004, 1041, 135–142
(17) Van der Klift, E. J. C.; Vivo´-Truyols, G.; Claassen, F. W.; Van Holthoon,
F. L.; Van Beek, T. A. J. Chromatogr., A 2008, 1178, 43–55
(18) Mottram, H. R.; Evershed, R. P. Tetrahedron Lett. 1996, 37, 8593–8596
(19) Mottram, H. R.; Woodbury, S. E.; Evershed, R. P. Rapid Commun. Mass
Spectrom. 1997, 11, 1240–1252
(20) Holcˇapek, M.; Jandera, P.; Zderadicˇka, P.; Hruba´, L. J. Chromatogr., A 2003,
1010, 195–215
(21) Byrdwell, W. C.; Neff, W. E. Rapid Commun. Mass Spectrom. 2002, 16,
300–319
(22) Fauconnot, L.; Hau, J.; Aeschlimann, J. M.; Fay, L. B.; Dionisi, F. Rapid
Commun. Mass Spectrom. 2004, 18, 218–224
(23) Leskinen, H.; Suomela, J. P.; Kallio, H. Rapid Commun. Mass Spectrom.
2007, 21, 2361–2373
(24) McAnoy, A. M.; Wu, C. C.; Murphy, R. C. J. Am. Soc. Mass Spectrom. 2005,
16, 1498–1509
(25) Cvacˇka, J.; Hovorka, O.; Jirosˇ, P.; Kindl, J.; Stra´nsky´, K.; Valterova´, I.
J. Chromatogr., A 2006, 1101, 226–237
(26) Malone, M.; Evans, J. J. Lipids 2004, 39, 273–284
(27) Iwasaki, Y.; Yasui, M.; Ishikawa, T.; Irimescu, R.; Hata, K.; Yamane, T.
J. Chromatogr., A 2001, 905, 111–118
.
.
.
.
.
.
.
.
.
.
.
(28) Marangoni, A. G.; Rousseau, D. In Food Lipids; Akoh, C. C., Min, D. B.,
Eds.; Marcel Dekker: New York, 2002; pp 301-333 and references therein.
.
3904 Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

Silver-Ion HPLC. HPLC was performed on a liquid chro-
matograph Agilent 1200 series (Agilent Technology, Waldbronn,
Germany). The final HPLC method for analyses of plant oils and
animal fats used the following conditions: three silver-ion chro-
matographic columns ChromSpher Lipids (250 mm × 4.6 mm, 5
µm, Varian, Palo Alto, CA) connected in series, the flow rate of 1
mL/min, the injection volume of 1 µL, column temperature of 25
°C, and the mobile phase gradient of 0 min 100% A, 140 min 61%
A + 39% B, where A is a mixture of hexane-2-propanol-acetonitrile
(99.8:0.1:0.1, v/v/v) and
B is the mixture of hexane-2-
propanol-acetonitrile (98:2:2, v/v/v). The mobile phase was
prepared fresh every day before analyses. Silver-ion columns were
conditioned at 50 µL/min of the initial mobile phase composition
overnight and at 1 mL/min for 1 h before the first analysis. The
injector needle was washed with the mobile phase before each
injection. The chromatographic system is equilibrated between
injections for 30 min. The hybrid quadrupole time-of-flight
(QqTOF) analyzer micrOTOF-Q (Bruker Daltonics, Bremen,
Germany) with positive-ion APCI was used in the mass range m/z
of 50-1200 with the following tuning parameters: flow of the
nebulizing and drying gas 5 and 3 L/min, respectively, temper-
atures of the drying gas and APCI heater 300 and 400 °C,
respectively. Reconstructed ion current chromatograms were used
to support the identification and quantitation of coeluting peaks.
Fatty Acid Abbreviations. M, myristic (C14:0); C15:0, pen-
tadecanoic; Po, palmitoleic (∆9-C16:1); P, palmitic (C16:0); Mo,
margaroleic (∆9-C17:1); Ma, margaric (C17:0); St, stearidonic
(∆6,9,12,15-C18:4); Ln, R-linolenic (∆9,12,15-C18:3); γLn, γ-lino-
lenic (∆6,9,12-C18:3); L, linoleic (∆9,12-C18:2); O, oleic (∆9-
C18:1); S, stearic (C18:0); C20:2, eicosadienoic (∆11,14-C20:2); G,
gadoleic (∆9-C20:1); A, arachidic (C20:0); B, behenic (C22:0);
C23:0, tricosanoic (C23:0); Lg, lignoceric (C24:0).
RESULTS AND DISCUSSION
Optimization of Silver-Ion HPLC of TGs. It is well-known
that the silver-ion chromatography suffers from a lower reproduc-
ibility of retention times in comparison to reversed-phase sys-
tems.29 In the present work, the maximum attention has been
paid to the optimization of chromatographic performance in terms
of regioselectivity, reproducibility, and robustness. The most
important parameters are the selection of column type and mobile
phase composition. The optimization of mobile phase composition
starts from the hexane-acetonitrile system which is the most
widespread system used in silver-ion HPLC13,30 providing the best
results concerning the regioisomeric resolution.12 The critical
problem of this frequently used mobile phase is a low mutual
miscibility of these solvents and the evaporative changes of
prepared mobile phases, which significantly contributes to the
reproducibility problems.29 This knowledge has driven us to test
the different mobile phase compositions with the goal to achieve
a better separation selectivity and reproducibility of retention times
for TGs. Extensive experiments with the combination of different
polar modifiers (2-propanol, ethanol, propionitrile, and ethylac-
Figure 1. Effect of separation temperature on silver-ion HPLC/APCI-
MS analysis of TGs in the randomization mixture of OOO/LLL: (A)
15, (B) 25, and (C) 40 °C. Other chromatographic conditions are
described in the Experimental Section.
etate) in hexane31 have shown that it is not easy to improve the
selectivity of the hexane-acetonitrile systems. Almost comparable
resolution is obtained with hexane-propionitrile systems, where
the reproducibility problems are not encountered (in agreement
with ref 32), but the serious drawback of this method is the toxicity
of propionitrile. We discourage the use of propionitrile because
of the health hazard in the laboratory. Other tested systems show
poor regioisomeric selectivity. On the basis of these experiments,
we have decided to add the third solvent to the hexane-acetonitrile
system to improve the mutual miscibility while keeping the
chromatographic performance, which has been achieved with the
hexane-acetonitrile-2-propanol system. Then, the optimization
(31) Vel´ınska´, H. Diploma Thesis, University of Pardubice, Pardubice, Czech
Republic, 2008.
(29) Harfmann, R. G.; Julka, S.; Cortes, H. J. J. Sep. Sci. 2008, 31, 915–920
.
(32) Mu¨ller, A.; Du¨sterloh, K.; Ringseis, R.; Ender, K.; Steinhart, H. J. Sep. Sci.
(30) Nikolova-Damyanova, B. Silver-ion HPLC of fatty acids and triacylglycerols.
In HPLC of Acyl Lipids; Lin, J.-T., McKeon, T. A., Eds.; HNB Publishing:
New York, 2005; pp 221-268.
2006, 29, 358–365
(33) Jakab, A.; Jablonkai, I.; Forga´cs, E. Rapid Commun. Mass Spectrom. 2003,
17, 2295–2302
.
.
Analytical Chemistry, Vol. 81, No. 10, May 15, 2009 3905

Figure 4. Silver-ion HPLC/APCI-MS analysis of randomization
mixtures: (A) PPP/LnLnLn; (B) PPP/OOO/LnLnLn.
Figure 2. Separation reproducibility for sunflower oil analysis: (A)
five consecutive injections in the first day; (B) five consecutive
injections in the second day. Chromatographic conditions are de-
scribed in the Experimental Section.
peak-broadening effects for multiple column coupling. On the basis
of these considerations, the comparison of one, two, and three 25
cm ChromSpher Lipids columns connected in series has been
performed (Supporting Information Figure S1). Retention times
are approximately doubled and tripled for coupling of two or three
columns but not exactly due to different retention characteristics
of individual commercial columns.
The chromatographic resolution of lipids in NARP systems is
significantly affected by temperature; generally, an improvement
is observed at lower temperature. Contrary to NARP chromato-
graphic systems, the retention in silver-ion HPLC increases with
increased temperature (Figure 1). Peak shapes at low temperature
(e.g., 15 °C in Figure 1A) are distorted. There is almost no visible
difference in the resolution between 25 °C (Figure 1B) and 40 °C
(Figure 1C), but the analysis time is shorter for 25 °C. Moreover,
40 °C is the maximum recommended temperature by the
manufacturer; hence, 25 °C is selected as optimum temperature
for further experiments.
Maximum attention has been paid to the right chromato-
graphic practice to reduce the fluctuation of retention times, which
involves mainly the following precautions. Mobile phases are
prepared every day fresh using solvents dried with molecular
sieves and kept in tightly closed bottles to avoid evaporation. All
measurements are performed in a single block of several weeks
on the same system without changing between normal- and
reversed-phase systems. Before any measurement, the columns
are conditioned using the low flow rate of initial gradient
composition (50 µL/min) overnight and the standard flow rate
for 1 h before the analysis.
Figure 3. Effect of the fatty acid chain length on the retention in
silver-ion HPLC/APCI-MS analysis of TGs in randomized sunflower
oil shown by reconstructed chromatograms of [XL]+ ions for XLL type
TGs, where X is saturated fatty acid from C15:0 to C24:0.
of gradient conditions has been performed31 resulting in the final
method described in the Experimental Section.
The commercial silver-ion ChromSpher Lipids column provides
the best performance and separation selectivity in the lipid analysis
based on our previous experiences and the literature data as
well.12,13,16,17,25 An increased length of chromatographic column
can improve the chromatographic resolution, as demonstrated
earlier in NARP20 or silver-ion13 separations of TGs. Unlike NARP
systems, the back pressure is not a limiting factor here, because
the mobile phases typically consist of hexane with a low percent-
age of polar modifier. The limiting factors are mainly long
retention times associated with the extended column length and
A remarkable improvement in the reproducibility of retention
times has been achieved in comparison to traditional hexane-
acetonitrile mobile phases previously used in our laboratory. The
3906 Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

Table 1. Basic APCI-MS Fragmentation Characteristics
of Analyzed Regioisomeric TGs in Randomization
Mixtures
Table 2. Peak Area Ratios of TG Regioisomeric Groups
in Silver-Ion HPLC/MS after the Randomization of
Equal Amounts of PPP, OOO, LLL, and LnLnLn
TG
DBa
1
ratio of [M + H - RiCOOH]+ ionsb
TG randomization
mixture
TG regioisomeric
groupa
peak area ratio
POP
OPP
[OP]+/[PP]+ ) 100:33
[OP]+/[PP]+ ) 100:80
OOO/PPP
OOP/OPO
OPP/POP
65:35
63:37
PLP
LPP
OOP
OPO
2
[LP]+/[PP]+ ) 100:39
[LP]+/[PP]+ ) 100:87
[OP]+/[OO]+ ) 100:62
[OP]+/[OO]+ ) 100:20
OOO/LLL
OOL/OLO
OLL/LOL
68:32
62:38
OOO/LnLnLn
LLL/PPP
OOLn/OLnO
OLnLn/LnOLn
66:34
65:35
PLnP
LnPP
OLP
LOP
OPL
3
[LnP]+/[PP]+ ) 100:49
[LnP]+/[PP]+ ) 100:97
[OL]+/[OP]+/[LP]+ ) 100:72:84
[OL]+/[OP]+/[LP]+ ) 100:97:53
[OL]+/[OP]+/[LP]+ ) 58:100:85
LLP/LPL
LPP/PLP
61:39
60:40
LLL/LnLnLn
LnLnLn/PPP
LLLn/LLnL
LLnLn/LnLLn
68:32
62:38
LLP
LPL
OLO
OOL
OLnP
LnOP
OPLn
4
5
6
[LP]+/[LL]+ ) 100:81
[LP]+/[LL]+ ) 100:33
[OL]+/[OO]+ ) 100:45
LnLnP/LnPLn
LnPP/PLnP
60:40
61:39
[OL]+/[OO]+ ) 100:88
[OLn]+/[OP]+/[LnP]+ ) 100:60:93
[OLn]+/[OP]+/[LnP]+ ) 95:100:33
[OLn]+/[OP]+/[LnP]+ ) 49:100:92
OOO/LnLnLn/PPP
OOO/LLL/PPP
OLnP/LnOP/OPLn
OLP/LOP/OPL
33:34:33
29:35:36
30:33:37
OLL
LOL
LLnP
LPLn
LnLP
OLnO
OOLn
[OL]+/[LL]+ ) 100:81
[OL]+/[LL]+ ) 100:34
OOO/LLL/LnLnLn
OLnL/OLLn/LOLn
[LLn]+/[LP]+/[LnP]+ ) 100:78:88
[LLn]+/[LP]+/[LnP]+ ) 44:100:82
[LLn]+/[LP]+/[LnP]+ ) 100:93:47
[OLn]+/[OO]+ ) 100:23
a sn-1 and sn-3 positions are not differentiated in this study, e.g.,
OOP and POO are treated as equivalent.
[OLn]+/[OO]+ ) 100:76
OLnL
OLLn
LOLn
LnLnP
LnPLn
[OL]+/[OLn]+/[LLn]+ ) 62:95:100
[OL]+/[OLn]+/[LLn]+ ) 94:54:100
[OL]+/[OLn]+/[LLn]+ ) 100:78:56
[PLn]+/[LnLn]+ ) 100:78
[PLn]+/[LnLn]+ ) 100:28
LLnL
7
8
[LLn]+/[LL]+ ) 100:36
[LLn]+/[LL]+ ) 100:68
[OLn]+/[LnLn]+ ) 100:79
[OLn]+/[LnLn]+ ) 100:34
LLLn
OLnLn
LnOLn
LLnLn
LnLLn
[LLn]+/[LnLn]+ ) 100:36
[LLn]+/[LnLn]+ ) 100:61
a DB, number of double bonds. b Listed values correspond to the
arithmetic mean of at least three consecutive measurements.
Figure 5. Positive-ion APCI mass spectra of (A) OLnP, (B) LnOP,
and (C) OPLn in a zoomed region of m/z 570-605 with diacylglycerol
fragment ions for peaks from the randomization mixture shown in
Figure 4B.
reproducibilities (Figure 2) for three selected peaks with 2 (PLP),
4 (PLL), and 6 (LLL) DBs are acceptable now, as demonstrated
on relative standard deviations of retention times of 0.4%, 1.0%,
and 0.7% for 1 day measurements and 1.9%, 1.7%, and 1.4% for 2
days measurements. In hexane-acetonitrile mobile phases used
in our laboratory previously, the 1 day reproducibility for these
three peaks was 7.4%, 6.8%, and 5.2%. Some shifts in retention times
can occur on longer time scale, but they can be efficiently
eliminated by the use of the relative retention r ) (tR,TG - tM)/
(tR,std - tM), as listed in Supporting Information Table S1. The
retention in silver-ion mode is governed mainly by the DB
number, but certain separation also occurs for TGs differing
only in the length of the fatty acid chain (see Figure 3). The
retention order of TGs within these groups with the constant
DB number in sunflower oil analysis is the following: for XLL
type (where X is the saturated fatty acid), LgLL < C23:0LL <
BLL < ALL < SLL < LLMa < LLP < LLC15:0 < LLM; for
XLO type, LgLO < C23:0LO < BLO < ALO < SLO < OLMa <
OLP; for XLS and XLP types, LgLS < BLS + LgLP < ALS +
BLP < SLS + ALP < SLP < PLP, etc. The reconstructed ion
current traces for diacylglycerol fragment ions of XLL series
are overlaid in Figure 3 to illustrate possible separation
according to the chain length. Nonlabeled minor peaks in this
figure correspond to mass interferences from other coeluting TGs
(not XLL series) with identical diacylglycerol fragment ions.
Retention characteristics of all TGs in particular DB groups
identified in plant oils (sunflower, blackcurrant, olive, and palm)
and animal fat (lard) are summarized in Supporting Information
Table S1. Differences in retention times of logical XLL series are
approximately 0.4 min per two methylene groups (see Figure 3).
The retention order of regioisomers follows the rule that more
DBs in sn-1/3 positions mean a stronger interaction with silver
ions embedded in the stationary phase and hence higher retention
in comparison with regioisomers having the same unsaturations
Analytical Chemistry, Vol. 81, No. 10, May 15, 2009 3907

regioisomers are generated for the optimization of regioisomeric
silver-ion separation. Chromatograms of randomization mixtures
(Figure 5 and Supporting Information Figures S2 and S3) also
demonstrate the stereoselectivity of our HPLC method providing
the baseline separation for regioisomers containing up to three
DBs and at least partial separation for TGs with four to eight DBs.
As a rule, the bigger the difference in the DB number of fatty
acids means better separation of corresponding TG regioisomers,
e.g., the baseline separation of P/L and P/Ln regioisomers is
relatively easily achieved (Figure 5 and Supporting Information
Figure S2). The example of partial resolution of LnOLn/OLnLn
regioisomers (Figure 4B) containing seven DBs shows that the
number of DBs is probably not a limiting factor for the regioiso-
meric separation, but the critical requirement for the successful
separation of highly polyunsaturated regioisomeric TGs is the
difference in the DB number between fatty acids. Polyunsaturated
TGs containing fatty acids differing only by one DB are critical
pairs for regioisomeric separation, e.g., O/L (peak splitting) and
L/Ln (only peak shoulder) pairs. Fatty acids differing by two and
more DBs are well-separated for TGs containing up to four DBs
and partially for five and more DBs. Some improvement may be
expected with further extension of column length at the expense
of very long retention times and also rather high costs of multiple
columns.
Figure 6. Silver-ion HPLC/APCI-MS analysis of (A) sunflower oil
APCI Mass Spectra of Regioisomeric TGs. The randomiza-
tion is used for the generation of regioisomeric mixtures of TGs,
which can be subsequently separated in silver-ion HPLC, and APCI
mass spectra of separated regioisomers can be recorded regard-
less the physical absence of pure regioisomeric standards. Figure
5 shows the differences in mass spectra of regioisomeric triplet
OLnP/LnOP/OPLn, which is chromatographically separated un-
like all previous papers. APCI mass spectra shown in Figure 5
correspond to the spectra of pure standards. The advantage of
silver-ion HPLC/MS determination of regioisomers is the fact that
regioisomers do not differ in their relative responses (see ratios
in Table 2); hence, peak area ratios correspond to concentration
ratios. Table 1 lists exact ratios of DG fragment ions obtained
only from randomization experiments to be sure that these results
are not affected by possible coelutions common in complex natural
samples. Two MS approaches are used for sn-2 fatty acid
characterization. The simple approach just determines the prevail-
ing fatty acid in the middle sn-2 position based on lower relative
abundance of the corresponding DG fragment ion,5,16,19,17,20,21,25
but it has been demonstrated recently34-37 that the type of fatty
acids (mainly DB number and positions) affects the relative
abundances of corresponding DG fragment ions. More exact
approach is based on the construction of calibration curves using
identical standards of regioisomeric pairs mixed at different
ratios,21-23,26,33 but the precision of this determination may be
affected by certain fluctuations of fragment ion ratios. Ratios of
DG fragment ions shown in Table 1 can be applied for NARP-
HPLC determination of regioisomeric ratios measured on the
same instrument, because the calibration curves are linear and
and (B) randomized sunflower oil.
in the middle sn-2 position. The probable explanation is a better
steric availability of the DB in sn-1/3 positions. Regioisomers of
XYO/YXO and XYL/YXL types (X and Y are saturated fatty acids)
and saturated TGs are not differentiated.
Randomization of TG Mixtures and Their Separation. The
chemical interesterification of fatty acids in TGs (randomization)
is a known process used in industrial processing of oils and fats
to modify physicochemical properties of food products.28 In our
work, the randomization is applied for the generation of standard
mixtures of regioisomers (Figure 4), which are too expensive or
often not commercially available at all. First, the randomization
reaction has been optimized31 resulting in a robust and reproduc-
ible procedure (see the Experimental Section for conditions).
Theoretically, the randomization of binary mixtures of equal
amounts of single-acid TGs R1R1R1 and R2R2R2 should provide
eight combinations of TGs at identical concentrations: R1R1R1,
R1R2R1, R1R1R2, R2R1R1, R1R2R2, R2R2R1, R2R1R2, and R2R2R2. In
practice, enantiomers R1R1R2 versus R2R1R1 and R1R2R2 versus
R2R2R1 cannot be resolved in a nonchiral environment; there-
fore, we obtain the following concentration ratios (in paren-
theses) for initial TGs and two regioisomeric pairs: R1R1R1 (1),
R1R2R1 (1), R1R1R2 + R2R1R1 (2), R1R2R2 + R2R2R1 (2), R2R1R2
(1), R2R2R2 (1). The theoretical calculation fits well with
experimental results for randomization reactions, as illustrated
in Figure 4A and Table 2 for binary randomization mixtures,
which confirms that the chemical interesterification is really a
random process applicable for the generation of regioisomeric
standards at defined concentration ratios. Figure 4B shows the
chromatogram of the ternary randomization mixture of OOO,
LnLnLn, and PPP providing six regioisomeric doublets at con-
centration ratios 2:1 and one triplet OLnP/LnOP/OPLn with
identical concentrations. This way the standard mixtures of TG
(34) Li, X.; Evans, J. J. Rapid Commun. Mass Spectrom. 2005, 19, 2528–2538
(35) Li, X.; Collins, E. J.; Evans, J. J. Rapid Commun. Mass Spectrom. 2006, 20,
171–177
(36) Gakwaya, R.; Li, X.; Wong, Y. L.; Chivukula, S.; Collins, E. J.; Evans, J. J.
Rapid Commun. Mass Spectrom. 2007, 21, 3262–3268
.
.
.
ˇ
(37) L´ısa, M.; Holcˇapek, M.; Rezanka, T.; Kaba´tova´, N. J. Chromatogr., A 2007,
1146, 67–77
.
3908 Analytical Chemistry, Vol. 81, No. 10, May 15, 2009

Table 3. Calculation of Theoretical TG Composition after the Randomization of TGs in Sunflower Oil (n Is the
Number of Fatty Acids, Ri Is Variable Fatty Acid, and xRi Is the Weight Fraction of Ri)a
no. of TGs
relative concentration of TGsb
sunflower oil
typeb
theoretical
sunflower oil
6
theoretical
3
xR
1
R1R1R1
n
LLL (23.3), OOO (1.2), PPP (0.05), SSS (0.01),
BBB (<0.01), AAA (<0.01)
R1R1R2
n(n - 1)
30
3xR 2xR
OLL (26.0), OOL (9.7), LLP (8.7), SLL (5.8), BLL
(1.3), OOP (1.2), LPP (1.1), SOO (0.80), ALL
(0.56), SSL (0.49), OPP (0.41), BOO (0.18), SSO
(0.18), SPP (0.09), AOO (0.08), SSP (0.06), BBL
(0.02), BPP (0.02), APP (0.01), BSS (0.01), BBO
(0.01), ASS (<0.01), BBP (<0.01), SBB (<0.01),
BBA (<0.01), AAO (<0.01), AAL (<0.01), AAP
(<0.01), AAS (<0.01), BAA (<0.01)
1
2
R1R2R3
total
[n(n - 1)(n - 2)]/6
[n(n + 1)(n + 2)]/6
20
56
6xR xR xR
3
OLP (6.5), SLO (4.4), SLP (1.5), BLO (0.97), SOP
(0.55), ALO (0.41), BLP (0.32), SLB (0.22), ALP
(0.14), BOP (0.12), ASL (0.09), BSO (0.08), AOP
(0.05), BSP (0.03), ASO (0.03), BAL (0.02), BAO
(0.01), ASP (0.01), BAP (<0.01), BAS (<0.01)
sum of TGs is 96.91%; the rest corresponds to TGs
containing trace fatty acids
1
2
a Analyzed sunflower oil has the following weight fractions in % of main fatty acids (ref 5): linoleic (L) 61.52%, oleic (O) 22.94%, palmitic (P)
7.69%, stearic (S) 5.15%, behenic (B) 1.14%, arachidic (A) 0.49%, and the rest (1.07%) corresponds to trace fatty acids. b Stereospecific configuration
is not distinguished in this table.
Table 4. Relative Peak Areas of Main TGs in Sunflower
Oil before the Randomization Compared with the
Experimental and Theoretical Relative Concentrations
after the Randomization
relative peak areas [%]
before
randomization
(exptl)
after
randomization
(exptl)
after
randomization
(theor)
TG
SLP
1.8
0.7
0.6
0.5
0.4
PLP
1.6
LPP/SOO
OOP
0/1.1
1.9
0.9/0.8
1.0
0.7/0.5
0.8
OSO
0.1
0.4
0.3
OPO
0.1
0.4
0.4
SLO
2.4
1.5
1.5
Figure 7. Silver-ion HPLC/APCI-MS analysis of lard.
OLP/SOL
OBL/LOP
OSL
2.0/4.3
0/3.1
0
1.8/2.7
0.4/2.5
1.4
1.5/2.2
0.3/2.2
1.5
Table 5. Comparison of Regioisomeric Occupation of
the sn-2 Position for Saturated (Palmitic) and
Unsaturated (Oleic and Linoleic) Fatty Acids in Plant
Oil (Sunflower) and Animal Fat (Lard)
OOO/OPL
BLL
3.7/0.1
1.6
1.5/2.7
0.9
1.2/2.2
0.9
SLL
7.5
4.1
3.9
LLP
9.7
5.8
5.8
OLO
3.0
3.6
3.2
regioisomeric pair
sunflower oil
lard
OOL
6.5
7.7
6.5
POP/OPP
OOP/OPO
PLP/LPP
100/0
98/2
8/92
LPL
0.4
3.0
2.9
12/88
1/99
OLL
16.4
4.9
16.0
9.4
17.3
8.7
100/0
97/3
LOL
LLP/LPL
9/91
LLL
17.2
20.0
23.3
OLP/LOP/OPL
63/36/1
3/12/85
the effect of mobile phase on the ratio of DG fragment ions is
negligible. This way the lack or high prices of pure regioisomers
can be overcome.
Information Figures S2 and S3). Table 3 shows the calculation of
theoretical TG composition in the randomized mixture28 applied
for the real sample of sunflower oil with known fatty acid
composition.5 To keep reasonable the complexity of this table,
only six main fatty acids are included in the calculation, while
remaining fatty acids form about 1%, and they are referred as trace
fatty acids. This calculation provides 56 combinations of TGs
(neglecting regioisomers in this table), but 27 of them have
negligible concentrations (<0.1%). The comparison of initial TG
composition of sunflower oil with experimental and theoretical
TG composition after the randomization is illustrated in Table 4.
Good correlation between theoretical and experimental values
Analysis of Complex Natural Mixtures. The randomization
process has been applied for sunflower, blackcurrant, olive, and
palm oils and lard as a representative of animal fats. Retention
times of all identified TGs in these samples are summarized in
Supporting Information Table S1. The comparison of chromato-
grams before and after randomization (Figure 6) is useful for the
verification of sn-2 determination, because the sn-2 occupation is
truly random after the interesterification, as shown in randomiza-
tion reactions of TG standard mixtures (Figure 5 and Supporting
Analytical Chemistry, Vol. 81, No. 10, May 15, 2009 3909

confirms that this model is applicable for the calculation of TG
concentrations after the randomization on the condition that the
initial fatty acid composition is correctly determined.
The optimization of HPLC conditions together with the coupling
of three 25 cm silver-ion ChromSpher Lipids columns in series
provides the best separation selectivity reported so far for the
regioisomeric analysis of TGs enabling the determination of
individual regioisomeric ratios for TGs found in natural samples.
In this work, our HPLC/APCI-MS has enabled the identification
of 196 TGs containing 0-11 DBs and fatty acid chain length from
14 to 24 carbon atoms. The comparison of representative plant
oil (sunflower) and animal fat (lard) provides quantitative data
(Table 5) for the preferential occupation of the sn-2 position by
unsaturated fatty acids in plant oils and saturated fatty acids in
animal fats. The randomization is used here as a new approach
for the generation of regioisomeric standards of TGs. The
randomization of selected plant oils shows that there is a clear
preference of certain fatty acid combinations and regioisomeric
order in TGs before the randomization process, whereas the TG
composition after the randomization is truly random. The position
of fatty acids on the glycerol skeleton is very important from the
nutrition point of view, and the presented method can contribute
to this area, because the absence of reliable quantitative methods
applicable for complex natural and biological samples limits such
research.
The generally accepted opinion is that unsaturated fatty acids
(mainly linoleic) preferentially occupy the sn-2 position in plant
oils, whereas it is just opposite for animal fats, where unsaturated
fatty acids are found mainly in sn-1/3 positions. The established
method for regiospecific determination is the enzymatic hydrolysis
of whole plant oils or animal fats, which provides overall informa-
tion on sn-2 average preference for all TGs found in this sample.
In our work, a representative plant oil (sunflowersFigure 6) and
animal fat (lardsFigure 7) are compared concerning the quantita-
tive data for sn-2 preferences of fatty acids in intact TGs. Table 5
lists regioisomeric ratios of TGs composed of three common fatty
acids (palmitic, oleic, and linoleic) occurring both in plant and
animal samples. This is a clear quantitative proof based on the
silver-ion separation of intact TGs that sn-2 occupation for saturated
versus unsaturated fatty acids is selective with opposite prefer-
ences for plants and animals. The interesting example is TG
composed of P, O, and L, where the sn-2 occupation preference
for sunflower oil is in the order of increasing DB number (OLP/
LOP/OPL ) 63/36/1), but it is just opposite for the lard sample
(OLP/LOP/OPL ) 3/12/85). Peak area ratios are obtained using
the reconstruction of suitable ion currents considering masses of
coeluting peaks. The data for other analyzed plant oils (palm, olive,
and blackcurrant oilssdata not shown) are in agreement with this
conclusion. The knowledge of sn-2 preference in TGs is very
important due to the bioavailability of particular fatty acids,
because human lipases first cleave fatty acids in sn-1/3 positions
and therefore fatty acids present in the sn-2 position can be worse
accessible for the human organism.
ACKNOWLEDGMENT
This work was supported by the Grant Project No. MSM-
0021627502 sponsored by the Ministry of Education, Youth and
Sports of the Czech Republic and Project Nos. 203/06/0219 and
203/09/P249 sponsored by the Czech Science Foundation.
SUPPORTING INFORMATION AVAILABLE
Additional information as noted in text. This material is
available free of charge via the Internet at http://pubs.acs.org.
Received for review January 21, 2009. Accepted March 23,
2009.
CONCLUSIONS
Silver-ion chromatography is a powerful method for the
analysis of TG regioisomers found in plant oils and animal fats.
AC900150J
3910 Analytical Chemistry, Vol. 81, No. 10, May 15, 2009