Journal of Chromatography, 435 (1988) 343-349

Elsevier Science Publishers B.V., Amsterdam - Printed in The Netherlands

CHROM. 20 106

CALIBRATION OF THIN-LAYER CHROMATOGRAPHY WITH FLAME

IONIZATION DETECTION FOR THE ANALYSIS OF NATURAL LIPID
SAMPLES

JANE F. WHITSETT and JOHN M. KENNISH*

Department of Chemistry, University qf Alaska-Anchorage, 3211 Providence Drive, Anchorage, AK 99508
(U.S.A.)
(First received July 30th, 1987; revised manuscript received September 28th, 1987)

SUMMARY

Thin-layer chromatography coupled with flame ionization detection has been

used to develop a method to quantitate fish lipids. Quantitation of fractionated lipid
classes is usually accomplished through calibration with standards. Our work shows
that various standards within a class, such as triglycerides or free fatty acids give
substantially different responses to the detector. A method has been developed in
which a composite sample of salmon lipid is prepared with an internal standard. This
technique eliminates the variable detector response observed for compounds within
a class and provides more accurate quantitation of oils from which the calibration
samples were prepared. The lipid classes quantitated were triglycerides, free fatty
acids, phosphatidylethanolamine and phosphatidylcholine.

INTRODUCTION

Separation and quantitation of complex lipids using thin-layer chromato-

graphy (TLC) with flame ionization detection (FID) has become an important tech-
nique’. This is largely due to its speed of analysis, simplicity and improved quanti-
tative capablities over conventional TLC. Usually, separation and quantitation of
lipids are achieved by TLC and various chemical methods, such as inorganic phos-
phorous determination after elution of the separated lipids from the TLC plate. This
method, among others is widely accepted and used. However, such methodology
involves extensive sample handling and transfer through various stages so that sig-
nificant losses of sample may occur. For TLC-FID the sample is spotted on a silica
gel-coated quartz rod, developed chromatographically, and finally detected by FID
as the rod is passed through a hydrogen flame. In this method sample handling is at
a minimum, but a new set of variables is introduced and requires further study to
fully utilize the method.
Several investigators have examined various aspects of lipid analysis with the
TLC-FIDzd5. In particular, it has been reported that response factors for various
lipids are different and depend upon variables such as sample volatility, amount of

OC21-9673/88/$03.50 0 1988 Elsevier Science Publishers B.V.

344 J. F. WHITSETT, J. M. KENNISH

material analyzed, fatty acid composition and the rate at which the Chromarod
passes through the flame+‘O.
Recently, Fraser et al.’ quantitated the neutral and polar lipids of the eggs and
larvae of marine fish using a composite standard of similar composition to the marine
fish. The impetus for this was that the response factor (ratio of analyte to internal
standard) of some lipids depends on the amount of sample present’ l. Because of this
it is important that the calibration solution consists of the analytes in the same pro-
portion as the biological sample. Results from this method were found to be com-
parable to the more common gravimetric, calorimetric and densitometric methods
for most analytes with the exception of triglyceride (TG), cholesteryl ester, phospha-
tidylethanolamine (PE) and phosphatidylcholine (PC). It has previously been demon-
strated by Kramer et ~1.~that the chromatographic behavior of lipid classes on Chro-
marods is influenced by the esterified fatty acids in the lipid class in that the RF value
of the lipid class increased with increasing chain length and degree of unsaturation.
The result is a variation in detector response such that response factors should be
determined for each lipid class with fatty acids resembling the lipid class to be ana-
lyzed9.
The method is very useful for the quantitation of entire classes of lipids such
as TG or free fatty acids (FFA). However, individual lipid standards, such as palmitic
acid or nervonic acid elicit different responses from FTD. This may be due to the
varying RF values of fatty acids and TG noted by Oshima et ~1.‘~ and Kramer et
aL8. Typically, a single standard is chosen for calibration of a given class. Natural
lipid samples, on the other hand, are comprised of a complex mixture so that a single
standard chosen from each class would not be representative. Thus, accurate quan-
titation requires elimination of the variables that are responsible for the observed
changes in detector response. We wish to report a novel calibration method, with its
application to lowering the error associated with changes in detector response for
lipids in the same class.

EXPERIMENTAL

Lipids were extracted from sockeye salmon muscle by the method of Radin12.
DEAE-Cellulose column chromatography was used to isolate the triglyceride fraction
according to Rouser et ~1.‘~. Calibration solutions were prepared by cornpositing a
given fraction from ten fish and adding pentacosane as internal standard. PE and PC
were quantitated using composites of whole lipid with internal standard added. Serial
dilutions were prepared from a 25 mg/ml stock solution. Internal standard concen-
tration was constant at 5 mg/ml throughout the dilutions.
A composite of whole lipid was hydrolyzed in alcoholic potassium hydroxide
for 2 h at 80°C. The hydrolysate was then acidified and extracted with hexane. This
extract was used to prepare free fatty acid calibration solutions in the same manner
as described above. Individual calibration solutions were prepared from palmitic
acid, stearic acid, arachidic acid, nervonic acid, triarachidonin, triolein and trimyris-
tin standards (Sigma).

Chromarod conditioning
SIX Chromarods were stored overnight in 4.5 A4 sulfuric acid solution. Before
TLC-FID OF NATURAL LIPID SAMPLES 345

use rods were rinsed 45 times in distilled water, dried at least 5 min at 1 10°C then
blank scanned once in the instrument. At the end of each day the rods were returned
to 4.5 M sulfuric acid for storage.

Thin-layer chromatography
Conditioned Chromarods were manually spotted with 1 ~1 of the sample so-
lution. The rods were air dried, then placed into a humidity chamber adjusted to
50% relative humidity with calcium chloride for 15 min. The Chromarods were then
transferred to a solvent chamber to equilibrate with developing solvent vapors for
15 min.
Neutral lipids were developed for 30 min in hexane-diethyl ether-formic acid
(80:25:1.2). After development the rods were dried for 5 min at 110°C. Polar lipids
were separated in chloroform-methanol-water (80:35:3.0) for 40 min, then dried as
above. Unfractionated salmon lipid samples were separated on Chromarods using
a double development scheme. Neutral lipids were developed, then partially scanned
leaving the polar lipids intact at the origin. Polar lipids were then separated and the
Chromarods fully scanned. Pentacosane, the internal standard, migrates to the sol-
vent front in the neutral lipid solvent and is used for quantitating both the neutral
and polar lipids.

Instrumental conditions
An Iatroscan (Model ThlO, Iatron Labs.) was operated at a hydrogen pressure
of 0.6 kg/cm2, an air flow of 2000 ml/min and a scan speed of 4.2 mm/s. A HP 3392A
integrator was used to record the chromatogram and individual peak areas.

RESULTS AND DISCUSSION

Calibration of the instrument using the fatty acid standards, palmitic, stearic,
arachidic, and nervonic acid and the fatty acid composite solution is shown in Fig.
1 and Table I. The data shown in Table I demonstrate the error associated with
quantitating a natural pool of fatty acids based on calibration with a single fatty
acid. The slope for the fatty acid composite calibration is 0.1234. The relative percent
difference (absolute value of difference between slopes, divided by the average of the
slopes, multiplied by 100) between this slope and that of palmitic acid is 14.0%. Of
the fatty acids examined palmitic acid (16:0) resembles the composite best though it
is statistically different. The relative percent differences for the slopes of the fatty acid
composite vs. standard are 17%, 32% and 60% for the fatty acids 18:0, 20:0 and
24: 1, respectively.
Apparently, the long chain mono-unsaturated fatty acid behaves differently as
it passes through the flame. In the case of 24:l the ionizable carbon (non-carbonyl
carbon) content is higher which has been cited as an explanation for the variable
detector response associated with TLC-FID3. Usually the response increases with
increasing non-carbonyl carbon atoms. Our results show the opposite effect. It is
possible that the compounds differ in their physical properties so that the kinetics
associated with the phase changes from a solid to a vapor and finally ions could be
sufficiently different to affect overall response.
Analysis of TG standards and the TG composite is shown in Fig. 2 and Table
II.
346 J. F. WHITSETT, J. M. KENNISH

TABLE I

TLC-FID CALIBRATION FOR FREE FATTY ACIDS

Standard r2 Slope C.V.* of slope (%)

16:0 0.9923 0.1073 12.5

18:O 0.9952 0.1037 9.85
20:o 0.9959 0.08894 9.10
241 0.9983 0.066607 5.92
(C.V. = 20.47%)

Fatty acid composite 0.9999 0.1234 1.2

l Coefficient of variation.

Data analysis of the triglycerides indicates that the variablity in the slopes
(C.V. = 17.5%) is too great to rely on a single TG standard for accurate quantitation
of that class. The slope for the triglyceride composite is 0.1380. The relative percent
difference between this slope and that of triarachidonin is 12.0% and is the closest
to the composite. For trimyristin and triolein the composite vs. standard relative
percent differences are 46.8% and 26.2% respectively. The chain length and unsatu-
ration of these compounds are sufficiently different that none of their responses are
comparable. Again, it is likely that sample volatility and ion formation are respon-
sible rather than the non-carbonyl content as for the fatty acids.
Kramer et ~1.~has related the peak shape to FID response in that sharp peaks
obtained from TGs with a single fatty acid composition gave a greater FID response
than mixed-fatty acid TGs. According to Kramer et al9 this is because the mixed
TGs spread over a larger area on the Chromarod resulting in a lower FID signal
since relatively more TGs are totally burned without ion formation. Our data, on the

CONCENTRATION mglml

Fig. I. TLGFID calibration. Curves: I = salmon lipid FFA composite; 2 = palmitic acid; 3 = stearic
acid; 4 = arachidic acid; 5 = nervonic acid.
TLC-FID OF NATURAL LlPID SAMPLES 347

0.0 4 I)- I
6 10 12 14 16 18 20 22

CONCENTRATION mglml

Fig. 2. TLC-FID calibration. Curves: I = salmon oil TG composite; 2 = triolein; 3 = triarachidonin;

4 = trimyristin.

other hand, show that the composites, prepared directly from the salmon lipid, elicit
a higher FID response. Perhaps the technique of exposing the Chromarods to the
developing solvent vapors before development allows the biological sample to mi-
grate in a band of similar width as a single fatty acid or single fatty acid triglyceride.
Also, the scan speed used (4.2 mm/s) is relatively fast which would minimize the
effect of zone broadening.
Several workers have quantitated the lipid classes investigated above14-17.
Quantitation is usually achieved using an internal standard and correction factors to
obtain the absolute weight of a given class. However, this does not correct for the
considerably different FID response to a mixture of molecular species, such as the
fatty acids esterified to the triglyceride pool in a biological sample. This aspect of
FID response to a class has been for the most part unnecessary since it has been used
primarily in gas chromatography where the response is due to a single molecular
structure. TLC on silica is often used for the separation of classes of compounds
rather than a single molecular species within the class as in gas chromatography’.

TABLE II

TLC-FID CALIBRATION FOR TRIGLYCERIDES

Standard Slope C. V.* of slope

Trimyristate 0.9993 0.08565 3.7%

Triolein 0.9985 0.1060 5.43%
Triarachidonin 0.9874 0.1224 16.0%
(C.V. = 17.5%)

TG composite 0.9893 0.1380 14.7%

l Coefficient of variation.
348 J. F. WHITSETT, J. M. KENNISH

TABLE III

LIPID CLASS COMPOSITION OF SOCKEYE SALMON MUSCLE

Lipid class Concentration (mg/g tissue)

f S.D. (n = 3)

TG 9.13 f 5.16
FFA 1.21 f 0.209
PE 0.388 f 0.123
PC 2.80 f 0.780

Therefore, the calibration sample must reflect as closely as possible the composition
of the natural sample of interest.
This approach has been applied to the analysis of TG, FFA, PE and PC in
freshly caught sockeye salmon stored under partial-freezing conditions (Table III).
The standard deviation associated with the values is primarily due to the variation
among the three fish rather than instrumental error. A chromatogram of an indi-
vidual fish stored at -20°C for 25 days is shown in Fig. 3.
In view of the points discussed so far the use of a composite calibration solution
prepared from the actual samples to be analyzed appears to be an improvement.
However, the analyst should consider the time invested in isolating the various frac-
tions. For analyses in which there are few samples it may be sufficient to calibrate
with a single fatty acid keeping in mind the error due to detector response. If several
samples are to be analyzed it may warrant the isolation of fractions and calibrating
with composites from them. In either case careful consideration of the events occur-

ISTD

SF

TIME min

Fig. 3. TLCFID analysis of sockeye salmon muscle lipid. Neutral lipids (left) were separated on type SII
Chromarods using the solvent hexane-diethyl ether-formic acid (80:25: I .2). ISTD = Pentacosane (inter-
nal standard); TG = triglyceride; FFA = free fatty acid; DC = diglyceride. Polar lipids (right) were
separated on type SII Chromarods using chloroform-methanol-water (80:35:3.0). PE = phosphatidyl-
ethanolamine; PC = phosphatidylcholine. SF = Solvent front.
TLC-FID OF NATURAL LIPID SAMPLES 349

ring as the Chromarod passes through the hydrogen flame is necessary. In the case
of the Chromarod, the positioning of the solid organic layer over the flame and under
the detector is not instantaneous, nor is the phase transition from solid to vapor and
then to ions. The calibration curves are not always a good linear fit as shown by their
correlation coefficients. This has been reported previously and it has been shown that
a polynomial regression produces a better fit4 which is unlike with gas chromato-
graphy-FID where linearity is easily demonstrated.
In conclusion, the TLC-FID method exhibits significantly different detector
response to individual compounds within a lipid class. The resulting quantitative
error can be diminished by calibrating the instrument with standards prepared from
the sample which is to be analyzed.

ACKNOWLEDGEMENT

This project was supported in part by the Alaska Seagrant College Program
under grant number NA86AA-DSG041, project number R/35-06.

REFERENCES

1 H. R. Harvey and J. S. Patton, Anal. Biochem., 116 (1981) 312.

2 S. M. Innis and M. T. Clandinin, J. Chromatogr., 205 (1981) 490.
3 R. G. Ackman, Methods Enzymol., 72 (1981) 205.
4 R. P. Delmas, C. C. Parrish and R. G. Ackman, Anal. Chem., 56 (1984) 1272.
5 E. W. Hammond, J. Chromatogr., 203 (1981) 397.
6 R. T. Crane, S. C. Goheen, E. C. Larkin and G. A. Rao, Lipids, 18 (1983) 74.
7 A. J. Fraser, D. R. Tocher and J. R. Sargent, J. Exp. Mar. Biol. Ecol., 88 (1985) 91.
8 J. K. G. Kramer, R. C. Fouchard and E. R. Farnworth, Lipids, 20 (1985) 617.
9 J. K. G. Kramer, B. K. Thomson and E. R. Farnworth, J. Chromafogr., 355 (1986) 221.
10 T. Oshima, W. M. N. Ratnayake and R. G. Ackman, J. Am. Oil Chem. Sot., 64 (1987) 219.
1I J. K. Kaitaranta and N. Nicolaides, J. Chromatogr., 205 (1981) 339.
12 N. S. Radin, Methods Enzymol., 72 (1981) 5.
13 G. Rouser, G. Kritchevsky, A. Yamamoto, G. Simon, C. Galli and A. J. Bauman, Methods Enzymol.,
14 (1969) 272.
14 J. R. Hazel, Lipids, 20 (1985) 516.
15 C. C. Parrish and R. G. Ackman, Lipids, 20 (1985) 521.
16 G. A. Rao, D. E. Riley and E. C. Larkin, Lipids, 20 (1985) 531.
17 J. K. G. Kramer, E. R. Farnworth and B. K. Thompson, Lipids, 20 (1985) 536.