Álvaro Suárez Jiménez – xwq109 - kucuris.ku.dk/ws/files/125297303/AlvaroSuarezJimenez.pdfMaster’s thesis Álvaro Suárez Jiménez – xwq109 High Field 1H-NMR analysis of pig
Post on 15-Jul-2020
0 Views
Preview:
Transcript
Master’s thesis
Álvaro Suárez Jiménez – xwq109
High Field 1H-NMR analysis of pig subcutaneous fat layers
Supervisor: Anders H. Karlsson
Co-supervisors: Flemming H. Larsen, Jorge Ruiz
Carrascal and Mette Christensen (Carometec a/s)
Faculty of Science – Department of Food Sciences
University of Copenhagen
2
TABLE OF CONTENT
1. Introduction Page 4
2. Theory background Page 5
2.1. Fatty acid profile and meat quality Page 5
2.2. Importance of rapid methods in the industry Page 8
2.3. Lipid extraction using balls mill method Page 10
2.4. HF-1H-NMR for fatty acid study Page 11
2.5. Iodine value Page 12
2.5.1. GC-FAME Page 13
2.5.2. NITFOM Page 14
3. Materials & methods Page 16
3.1. Animals and sample preparation Page 16
3.2. HF-1H-NMR Page 17
3.3. Fat extraction Page 18
3.4. Fatty acid composition (FAME) Page 18
3.5. Iodine value Page 20
3.6. Data analysis Page 20
4. Results & discussion Page 20
4.1. Gas chromatography analysis Page 21
4.2. HF-1H-NMR analysis Page 25
3
5. Conclusion Page 31
6. Acknowledgements Page 31
7. Literature Page 32
4
1. INTRODUCTION
The aim of the study was evaluate the importance of the different
subcutaneous fat layers for iodine value calculation on Danish pigs. A
previous study by the company CAROMETEC where the iodine value was
calculated through the subcutaneous fat without layer differentiation was used
as reference material. The analysis was performed using GC-MS and
HF-1H-NMR and different data analysis tools were used to evaluate the
calculation established. A clear differentiation between fatty acids composition
between layers was detected where bigger amount of unsaturated fatty acids
was found in the external subcutaneous fat layer and more saturated in the
internal layer. More accurate calculations for iodine value were developed
using the information from the inner fat layer.
5
2. THEORY BACKGROUND
2.1. Fatty acid profile and meat quality
Meat quality is defined as a combination of properties, including
technological quality attributes, consumer acceptance, and credence
characteristics of safety and health, as well as more intangible features such
as the cleaning, green or welfare status of the production system (Lee et. Al.
2014). The consumer increasingly prefers products with a higher unsaturated
fatty acid composition because of their beneficial effects in preventing
cardiovascular diseases. Pig diets supplemented with vegetable oils such as
soybean oil, sunflower oil, and corn oil, contain a high percentage of
unsaturated fatty acids and should lead to healthy products for consumers.
(Mitchaothai et. Al. 2007).
In meat and in related products the total fatty acid content strongly
influences the physical-chemical characteristics of foods, such as elasticity,
texture, mouthfeel, juiciness and lubricity (Siliciano et. Al. 2013).
Diet influenced growth rate and fatness, the low protein diet slowing
growth and producing fatter meat, more so in the two modern breeds, and
particularly in intramuscular rather than subcutaneous fat. This diet produced
more tender and juicy meat, although pork flavour and flavour liking were
reduced (Wood et. Al. 2004). In non-ruminants, the fatty acid pattern of dietary
lipids is reflected in the fatty acid composition of tissues (Mitchaothai et. Al.
2007).
The major public health institutes and different authorities,
independently, recommend a daily balanced proportion of saturated,
monounsaturated, and polyunsaturated fatty acids in the diet for a correct
nutrition and a healthy lifestyle. In western industrialized countries the current
indications for lipid intake have raised the question of the nature of fatty acid
effects on human health (Siliciano et. Al. 2013). Although beneficial health
effects of an increased intake of polyunsaturated fatty acids was noted, it is
still not definitively assessed, and negative effects of n-6 PUFA have neither
6
been established (Siliciano et. Al. 2013). The diet with high polyunsaturated
fatty acids resulted in higher levels of polyunsaturated fatty acids in back fat
and more rancid loin and sausage products after one and eight months of
freezer storage compared to those from the low PUFA diet (Bryhni et. Al.
2002). This trend of increasing unsaturated fats on meat products involve
several negative effects on meat and carcass quality, such as soft adipose
tissue, difficult slicing, higher susceptibility to lipid oxidation with the
generation of toxic reactive compounds (Martin et. Al. 2007).
Differences between intramuscular and subcutaneous fat has been
studied to be able to relate fat quality of both tissues. The positive effect of
age on intra muscular fat cannot be transpose to subcutaneous where a
negative trend has been found in experiments in which pigs reached heavy
weights (Bosch et. Al. 2012). It has been also found differences in the fatty
acid composition of the outer and inner subcutaneous back fat layers from
selected pigs showing that the outer layer is more unsaturated than the inner
layer (Daza et. Al. 2007).
Total saturated fatty acids decreased from inside to outside the back-
fat, being higher in the inner and showing the lowest proportion in the outer
one. Total monounsaturated fatty acids decreased from outside to inside, the
highest proportion being that of the outer layer, and the trend for total
polyunsaturated fatty acids was similar to that of monounsaturated fatty acids,
higher levels outside and lower inside (Daza et. Al. 2007).
Overall, significant differences in the fatty acid profile of the three studied
subcutaneous fat layers were reported, with a general trend to a higher
unsaturation when the layer is closer to the animal outer surface. Such
variations could be partially due to a different metabolism in each layer, aimed
to keep the fat fluid at ambient temperature (unsaturated fats have lower
melting temperature than saturated fats) (Daza et. Al. 2007). Fat sources with
high proportions of polyunsaturated fatty acids or monounsaturated fatty acids
are commonly used or at least have been tested for swine feeding. These
feeding strategies allow a substantial modification of the fatty acid profile of
pork meat and meat products, leading to more unsaturated fats, which follows
7
health advice about the decrease in the consumption of saturated fats due to
their implication in cardiovascular diseases (Martin et. Al. 2007).
Nevertheless, it seems that the outer layer is the first one deposited, followed
by the inner, and thus, variations in the fat content and the fatty acid
composition of the feeding sources during the different phases of the rearing
system could be also implied in the overall fatty acid profile of each layer
(Daza et. Al. 2007).
The increase in saturated fatty acids and the decrease in
monounsaturated fatty acids and polyunsaturated fatty acids contents of
subcutaneous adipose tissue of conjugated linoleic acid fed pigs have been
also found. The inhibition of the ∆9 desaturase by conjugated linoleic acid has
been suggested as the main reason explaining the modifications in total
saturated fatty acids and monounsaturated fatty acids contents obtained in
most of the studies. On the other hand, the inhibitory effect of dietary
conjugated linoleic acids on other desaturase activities could be also the
reason explaining the observed decrease in some polyunsaturated fatty acids
contents (Martin et. Al. 2007).
Monounsaturated fatty acids supplementation significantly affected the
contents of total saturated fatty acids and monounsaturated fatty acids of
subcutaneous adipose tissue. Thus, the back fat from pigs fed the high-
monounsaturated fatty acids experimental diets reached the highest
proportions of monounsaturated fatty acids and the lowest of saturated fatty
acids (40.35% of monounsaturated fatty acids and 39.04% of saturated fatty
acids, average value regardless of conjugated linoleic acids level), whereas
those animals fed the low monounsaturated fatty acids treatments showed the
lowest monounsaturated fatty acids and the highest saturated fatty acids
contents (37.74% of monounsaturated fatty acids and 41.37% of saturated
fatty acids, average value regardless of conjugated linoleic acids level) (Martin
et. Al. 2007).
The inclusion of oils with a high proportion of C18:2 in the feeding
increases the concentration of C18:2 in the tissues and reduces the
concentration of endogenous synthesized fatty acids (monounsaturated fatty
8
acids and saturated fatty acids), whereas the use of a high C18:1 content
feeding enhances the proportion of C18:1 in pig tissues, decreasing the
amount of saturated fatty acids, although polyunsaturated fatty acids content
is not significantly modified (Ruiz and Lopez-Bote, improvement dry cured
ham). According to Bryhni et. Al. (2002) it is emphasize the importance of
controlling the PUFA content in feed to avoid problems during storage and
processing.
As shown for the number of studies focus on this topic, the relation
between feeding and the fatty acid profile presented on the meat product is
increasing its importance. The rising significance of healthy products for
consumers is motivating bigger efforts to offer more beneficial meat for health
on the market.
2.2. Importance of new rapid methods in the meat
industry
One of the important quality parameters in porcine carcass grading for
determining farmer payment and carcass sorting before splitting and cutting is
the quality of the fat in the porcine carcass, as it is important for many
parameters of meat quality (Viereck, Sørensen and Engelsen, 2012). Classic
analysis methods for fat analysis have big reactive expenses and they are
time-consuming.
Another important fact when speaking about analysis of raw materials in
the meat sector is the big challenge due to huge range of variability of the
incoming meat. This high variability turns into high variability in quality of the
products and smaller control of the processes on the production lines and final
products. Thus, analysis methods for using on the meat sector needs to be
able to cover this huge variability.
The classic method for analysis of fatty acid profile of food is gas
chromatography. This method shows good accuracy but it requires extensive
sampling and is time-consuming (Dalitz et. Al. 2012). Hence, new methods
9
have been investigated to find a proper method which could be used at on-
line process. The high speed of processes on the meat industries (up to 1200
carcasses per hour) makes spectroscopy methods the most factible technique
for this aim.
According to Prieto et. Al. (2009) visible and near infrared reflectance
spectroscopy (Vis-NIR) has the potential of predicting quickly and accurately
different attributes of meat quality and it is suitable for on-line use and for
simultaneous determination of different traits. Because of these advantages,
the technology is being broadly used by the industry research-base for large-
scale meat quality evaluation to predict the chemical composition of meat
(Prieto et. Al. 2009). The main inconvenience for spectroscopy methods is the
need of very robust models capable of covering the big range of variability of
raw materials analyzed. Some characteristics, which could be studied with
rapid methods, are important criteria that affect consumers’ evaluation of meat
quality. Hence, there is an urgent need to find a fast and efficient alternative
method to estimate these criteria (Prieto et. Al. 2009).
Focusing on NMR methods, low field NMR has been displayed as a
appropriate method for water/moisture analysis on fat (Todt et. Al. 2006).
Thus, fraud by addition of water could be found by rapid NMR analysis. 1H
high resolution NMR has also given some interesting results in meat
authentication and determination of geographical origin. Common adulteration
practices consist both in undeclared mixing of meat from different species and
in mixing of expensive with cheaper meat (Mannina et. Al. 2012).
Regarding iodine value, an early prediction enables the slaughter
industry to quantify fat quality and thereby use the information for product
sorting resulting in increased production efficiency and economic gain.
10
2.3. Lipid extraction using balls milI method
One of the most time consuming steps on the fat analysis methods is the
lipid extraction. Efforts to achieve a method is present in some studies (Perez-
Palacios et. Al.2008).
According to Perez-Palacios et. Al. (2008), chemical and physical
treatments used for lipid extraction must remove them from their binding sites
with cell membranes, lipoproteins and glycolipids. It has been demonstrated
that the use of different methods results in different lipid recoveries in
biological samples. Indeed, results varied widely due to differences in
extraction methodology.
The standard methodology has been for decades solid–liquid extraction
procedures. These are based on a solvent (hydrophilic or hydrophobic, acidic,
neutral or basic) added to a solid. Insoluble material can be separated by
gravity or vacuum filtration, and soluble material is ‘extracted’ into the solvent
(Segura & Lopez-Bote 2014). The search for new and accurate lipid
extractions methods in meat and meat products is a very demanded topic. In
fact, depending on the tissue source, multiple groups adapted the
conventional protocols or developed new ones (Segura & Lopez-Bote 2014).
Although pure lipids are soluble in a wide range of organic solvents, the
model solvent or solvent mixture for extracting lipids should be polar enough
to remove such lipids from their association with cell membranes and tissue
constituents, but also not so polar that the solvent does not readily dissolve all
triacylglycerol molecules and other nonpolar lipids and, of course, should not
react chemically with the extract (Segura & Lopez-Bote 2014).
It has been shown that the balls mill protocol (which it was the extraction
method on this study) allows the analyst to treat a large number of samples in
a shorter time than the classic FOL method. Attending to fat content, the balls
mill mehotd offers analogous results to FOL method in quantity, but it shows
lower variability. In case that the fatty acid profile was the pursued goal, OS
extraction (where extract and methylate fatty acids is done in a Toluene-
Methanol/HCl solution in a rapid one-step procedure) showed a similar speed
11
than balls mill method but results obtained with balls mill are closest to those
obtained with the classic FOL procedure. (Segura & Lopez-Bote 2014)
2.4. HF-1H-NMR for fatty acid study
The perfect process analytical method would be based on a robust, non-
invasive and easy to handle customized technique operating in real-time. The
ideal instrumentation comes without any need for calibration (absolute
method), it has a professional support, and it is compliant to increasing
regulatory requirements. At least for NMR a trend can be observed towards
such an all-in-one device suitable for every purpose (Dalitz et. Al. 2012).
High-resolution 1H-NMR has been proposed as a fast and accurate
technique suitable for the analysis of edible oils and fats. This spectroscopic
method, often accompanied by the application of very simple mathematical
systems of calculation, has successfully been employed for the analysis of
acyl chain composition in olive oil and other edible vegetable oils, providing
huge information about their use in human nutrition. High-resolution NMR
spectroscopy, complementary to chemometric analysis, has been used as a
rapid quality control and authentication tool (Siliciano et. Al. 2013).
Quantitative high-resolution on-line NMR spectroscopy can be applied to
the investigation of complex fluid mixtures containing analytically similar
compounds. The development of on-line (flow) techniques has tremendously
increased the value of NMR spectroscopy as a non-invasive method for
process development applications (Dalitz et. Al. 2012). The application of this
kind of spectroscopy to the analysis of animal fats is restricted. At this time,
and at the best of our knowledge, meat and meat products apparently seem
to be seldom explored (Siliciano et. Al. 2013). However, the usefulness of 1H
NMR spectroscopy has been increasingly recognized for its noninvasiveness,
rapidity and sensitivity to a wide range of compounds in one single
measurement (Christophoridou and Dais 2009).
It is possible to settle that the routine application of high-resolution 1H
12
NMR techniques in the screening of the quantitative composition of fatty acid
chains is a useful addition to the methods for the analysis of the lipid fractions
of meat products. The short time required for sample preparation and spectra
recording can allow the analysis of a larger number of matrices per day. High-
resolution 1H-NMR spectroscopy could be used as a powerful tool in
alternative, to the classical chromatographic methods for the determination of
fatty acid chain compositions in meat products (Siliciano et. Al. 2013). In those
cases where liquid state is necessary, deuterated chloroform is an appropriate
chemical compound. On the deuterated chloroform the H from the chloroform
molecules is replaced by a deuterium isotope. This avoid the signal from the
H on the chloroform improving the received signal.
As described by Viereck, Sørensen and Engelsen (2012), high
resolution NMR has been demonstrated as suitable for subcutaneous fat layer
differentiation directly on fat tissue. The PLS model developed on Viereck,
Sørensen and Engelsen (2012) study could to some extend predict iodine
value with a R2 of 0.71 and a prediction error RMSECV of 2.11 using 5PCs.
The possibility of improving this prediction using liquid state samples was
developed in the present research.
2.5. Iodine value
Iodine value is a measure of the average amount of unsaturation of fats
and oils and is expressed in terms of the number of centigrams of iodine
absorbed per gram of sample (% iodine absorbed). The iodine value can be
determined using a traditional titration, or by gas chromatographic
quantification, however, both methods are quite time consuming. Hence,
spectroscopic techniques have been shown to be useful as faster methods
(Viereck, Sørensen and Engelsen, 2012).
Iodine value is used as a quality indicator for several parameters on the
meat industry. The range of information related to the iodine value is width,
from Information about rearing system of the animals to information to choose
the best technological use of the raw material. In the other hand, it is
13
important to consider that iodine value is only an indicator of number of
unsaturations, but it does not give any information about the kind mollecules
where the double bonds are presented.
The amount of unsaturation of the constituent fatty acids has been
measured by the iodine value, which is currently determined by the Wijs
method. Major drawbacks of that method include the use of dangerous iodine
dichloride (Wijs reagent) and the time-consuming procedures for reagent
preparation and chemical analysis. A procedure to determine the iodine value
from the fatty acid composition has been proposed by an American Oil
Chemists’ Society (AOCS) method (Kyriakidis and Katsiloulis 2000).
Accordingly, the proposed procedure for the calculation of the iodine value
from the percentages of fatty acid methyl esters (FAME) by and coefficients
specific for every vegetable oil can be used successfully for the determination
of iodine value determination (Kyriakidis and Katsiloulis 2000).
2.5.1. GC-FAME for iodine value calculation:
Iodine values were calculated through the method use for the American
Oil Chemists’ Society. Gas chromatography analysis of the fatty acid methyl
esters is performed on the samples and once the concentration of individual
fatty acids is known, through calculation of the following formula, iodine value
is expressed:
Iodine value = (% 16:1 × 0.950) + (% 18:1 × 0.860) + (% 18:2 × 1.732) +
(% 18:3 × 2.616) + (% 20:1 × 0.785) + (% 22:1 × 0.723)
This method is a good improvement regarding the Wijs method, but it is
still quite time-consuming and it is not possible to use it in a normal meat
industry line speed. Anyhow, it is the most appropriate method for calibration
of spectroscopy techniques.
14
2.5.2. NITFOM:
NitFom is an invasive handheld measuring device, which predicts the
iodine value and fatty acid profile of pork fat with a measurement cycle of 3
seconds.
This technique has been developed by the company CAROMETEC in
co-operation with the Danish National Advance Technology Foundation and
the University of Copenhagen (Faculty of Science, Department of Food
Science). This system uses Near-Infrared-Transmission spectroscopy in
combination with chemometric modeling for predicting several fat
characteristics as iodine value, or melting point.
Figure 1. NITFOM equipment (From CAROMETEC)
With a short measurement cycle (3 seconds), this equipment fits perfect
on the high speed production lines common on the meat industry. The
NITFOM can be used to get information pre-measurement and post-
measurement. Thereby, feed regimen of the animal can be known, as product
quality or slicing yield optimization. The iodine value is calculated with a
precision of ± 2.0 on cold carcass.
The use companies are giving to the equipment is wide depending on
the country. While in USA the companies are more interesting on get
information pre-measuremente about the feed regimen of the animal
slaughtered, on Germany the method is being used to get more complete
information about the requirements for the subsequently cooking hams.
The measurement (1100-2200 nm) is performed on the fat tissue in the
15
neck region where the probe penetrates 4 cm into tissue. The speed with
which the NitFom probe is retracted from the carcass determines how many
spectra are obtained (8 measurements on 3 seconds).
Figure 2. Applied spectroscopy of the NITFOM equipment (From CAROMETEC)
Model is built using PLS and using as reference values a GC analysis
performed by the Danish Meat Research Institute. On hot carcass
classification R2cv was 0.94 with a RMSECV = 1.8 IV, while on cold carcass
R2cv was 0.93 and RMSECV = 1.8 IV.
16
3. MATERIALS & METHODS
3.1. Animals and sample preparation
A total of 90 samples of subcutaneous fat from pig, from skin to meat
presented were obtained from a local slaughterhouse by the company
CAROMETEC. The animals were chosen among normal Danish slaughter
pigs breed including extreme samples regarding their iodine values (from
approx. 60 to 80). Pigs were slaughtered on March 26th 2014. From the
company CAROMETEC we got estimated iodine values measured on 49 of
these samples through two different methods described previously (GC-FAME
and NITFOM). These iodine values were calculated based on the whole fat
piece on both methods without layer differentiation (both fat layers from the
samples were melted together before analysis). Measurements with the
NITFOM-instrument were performed on cold carcasses (after one night
storage at a temperature around 6 ºC). Sample size was approx. 5x2 cm
times 6 cm deep and were numbered from 10 to 99.
The samples were kept on -18 ºC until the analysis was performed.
Samples were divided manually in frozen state with a knife into 5 different
layers. These layers are from outside to inside the animal:
-‐ Skin
-‐ Upper or external subcutaneous fat layer
-‐ Link between both subcutaneous fat layers
-‐ Lower or internal subcutaneous fat layer
-‐ Muscle layer
The differentiation of the 4 upper layers were not completely clear in
some of the samples due to the irregular shapes of the samples or the non
parallel line of the link between fat layers close to the skin. Besides, several
samples show brown dots on its external fat layer (Figure 3), probably due to
some kind of skin penetrations on the external fat layer, but as it will be shown
later, it did not affects the results on the fatty acid analysis. Skin, connective
tissue between both layers and muscle were discarded, while the two different
fat layers were selected for the upcoming experiment. They were named U for
17
the external layer, and D for the internal layer, plus the previous mentioned
numbers of the sample. Thereby, in total, 98 samples were obtained.
Figure 3. Sample U26 before homogenization.
Subsequently, after separating the two layers, each layer was
homogenized. For this homogenization, samples were frozen in liquid nitrogen
and minced using a Moulinex . After mincing, the samples were frozen and
stored for further analysis at -18 ºC. Thereby, 98 samples where 49 belong to
the external fat layer and 49 to the internal fat layer. These samples were later
analyzed by GC-MS and HF-NMR methods.
3.2. HF-1H-NMR analysis
To ensure an appropriate signal on the future NMR analysis, the biggest
amount of fat dissolved without presenting precipitation was chosen. A test to
calibrate this optimal dissolution of fat in CHCl3 was carried out using 600 µL
of chloroform as solvent and different amounts of fat for testing. Optimal
amount was selected as 50 mg. The samples were Vortex mixed (30
seconds) and centrifuged afterwards (10 minutes at 5000 rpm). For the HF-1H-NMR experiment 250 µL of the dissolution previously created was mixed
with 250 µL of deuterated chloroform.
Spectra of the 98 samples were collected using a Bruker Avance DRX-
500 (Bruker BioSpin, Theinstetten, Germany) spectrometer, operating at a
frequency of 500.13 MHz for protons equipped with a double-tuned BBI
probe. As described before, 50% of CDCl3 was added as lock solvent.
Diameter of sample tube was 5 mm. Data were recorded at 298 K. Other
recording parameters include 32k data points, 64 scans, recycle delay of 5 s,
a spectral width of 10000 Hz and an acquisition time of 1.639 s. Spectra was
aligned according to the (CH2)n resonance (approximately 1.25 ppm) of the
18
fatty acids. After alignment only the spectral region 0.5-6.0 ppm was used for
further analysis.
3.3. Fat extraction
The fat extraction for the GC-MS analysis was performed using the
method described by Segura and Lopez-Bote (2014). In a 2 mL safe-lock
micro test tube, 200 mg of the homogenized fat sample was accurately
weighed. Four steel balls (2 mm Ø) and 1.5 mL of CHCl3:MeOH 8:2 mixture
was added. After being tightly capped, the tubes were placed in an adapters
and was homogenized for 2 min at 20 Hz in a Mill MM400 mixer (Retsch
technology, Haan, Germany). The resulted biphasic system was allowed to
separate by centrifugation (8 min, 10,000 rpm, 25ºC). The solvent was
decanted into a previously weighted 4 mL vial and thereafter stored in an
freezer at -4 ºC.
3.4. Fatty acid composition (FAME)
The fatty acid composition was determined according to Jart (1997) with
some modifications: 100 µL of extraction solvent were transferred to a 10 mL
test tube. Solvent was evaporated under nitrogen stream at 25 ºC. 1.00 mL
sodium-methylate solution was added, followed by a Vortex mix for 30
seconds. The test tubes were placed in a 60 ºC water bath until the samples
were clear (after this the lipid phase was not visible anymore). The samples
were kept in the water bath for an additional 30 minutes. When samples were
removed from the bath 4.00 mL of saturated sodium chloride was added and
1.00 mL of hexane, and thereafter it was Vortex mix for 30 seconds. The
samples were left until the phases have separated. The top phase is the
hexane and contains the methylated fatty acids, the samples were evaporated
and dissolved in 1.0 mL of hexane prior to analysis on the gas chromatograph
(HP 6890 series, Hewlett-Packard, Palo Alto, CA, USA) with a 30 m x 0.32
mm x 0.25 µm Omegawax column (Supelco, Bellefonte, USA) and with FID
detection. The oven temperature programmed and conditions were: 50 ºC for
1 minute; from 50 to 180 at 15 ºC/minute; from 180 to 240 at 3 ºC/minute; at
240 ºC, held for 10 minutes. Injected volume was 1 µL and the split ratio was
1:25. Hydrogen was used as carrier gas and the flow was 1 mL/minute. The
19
results were analyzed with Chemstation software (Agilent Technologies) and
the fatty acid methyl esters were identified by comparing retentions times with
known standards. Fatty acid presented in most of the samples were:
· Myristic acid à CH3(CH2)12COOH à C14:0
· Palmitic acid à CH3(CH2)14COOH à C16:0
· Palmitoleic acid à CH3(CH2)5CH=CH(CH2)7COOH à C16:1cis-Δ9
· Stearic acid à CH3(CH2)16COOH à C18:0
· Oleic acid à CH3(CH2)7CH=CH(CH2)7COOH à C18:1cis-Δ9
· Linoleic acid à CH3(CH2)4CH=CHCH2CH=CH(CH2)7COOH à à C18:2cis-Δ9, Δ12
· α-Linolenic acid à CH3CH2CH=CHCH2CH=CHCH2CH=CH(CH2)7COOH à
à C18:3cis-Δ9, Δ12, Δ15
The results were reported as % fatty acid of the total content of detected
fatty acids. Duplicates for the fatty acid analysis through GC were carried out
performing the methylation and analysis on the gas chromatographer two
times.
Results from the GC-MS on paper transferred manually to a computer.
As it is shown on Figure 4, the results from the GC-MS shows some ambiguity
close to the retention time belonging to C16:0. Due to similar reasons several
samples were not completely clear on the transcription. However, the most
consistent possibility was chose supporting the decisions with the literature.
Figure 4. GC results 2nd replicate of sample U58
20
3.5. Iodine value
Two calculated iodine values given by CAROMETEC were used on the
data analysis (one calculated through GC-FAME analysis and the other value
calculated through the NITFOM), and a third iodine value was calculated
based on the formula from the AOCS method for iodine value calculation
using GC-FAME.
3.6. Data analysis
Multivariate data analysis for GC-MS and HF-1H-NMR was performed in
the form of principal component analysis (PCA) and partial least squares
regression (PLS) to obtain optimal quantitative and qualitative information
from the measured experimental spectra. Different pre-processing techniques
were necessary for both methods. 18028 data points were resulted for
modeling of the HF-1H-NMR analysis. All models were validated using Full
cross-validation, from which the root mean square error of cross-validation
(RMSECV) was calculated to measure prediction error. Data was analyzed
using Microsoft Excel and the chemometric software LatentiX 2.12
(www.latentix.com, Latent5, Copenhagen, Denmark).
4. RESULTS & DISCUSSION
As first step, the average value of the fatty acid composition based on
the two replicates was calculated for each of the 98 samples. From here on,
the rest of the analyses were based on this average.
It is remarkable the non detection of some FA in several samples:
Oleic acid (16:1): Not detected in samples U14 and U99.
Linoleic acid (18:2): Not detected in sample U51.
Alfa-linoleic acid (18:3): Presented in 77.5% of the samples.
21
4.1. Gas Chromatography analysis
Table 1 shows simple statistical analyses of the results obtain from the
GC-MS based on the average of both replicates. Coefficient of variation (CV)
shown for both layers, could have as reason the differences on fatty acid
composition between layers, nonetheless the CV in each layer is a good
explanation of the precision achieve on the study. As it is presented the
internal fat layer achieve better results than the external and the reason
explaining this fact is not only regarding the probability of small errors on the
analysis. Regarding Daza et. Al. (2007) the external layer is the first one were
the new fat is deposited, followed by the internal one. The older presence of
fatty acid in the body could explain the bigger stability of the compounds
presented, thus increasing the variability of the composition in the external
layer. Anyway, excessive coefficients of variation are shown on fatty acids
16:1 and 18:0 of the external layer as in 18:3 of both layers. The high CV
value on 18:3 could be explain as well due to the small amount of this fatty
acid on the samples where is on average 0.93%.
Table 1. Content of fatty acid on the study where upper is external and lower internal fat layer
The iodine values provided by CAROMETEC were calculated through
two different methods: GC-FAME and NITFOM. As it is shown on Figure 5a
the correlation between both methods was really good (R2 = 0,95355). The R2
between the iodine value calculated and the iodine value receive from
CAROMETEC was 0.84 (Figure 5b).
22
Figure 5a. Correlation between iodine value calculated with GC-‐FAME and iodine value calculated using NITFOM
Figure 5b. Correlation between iodine value calculated and Iodine value GC-‐FAME from CAROMETEC
The first step on the data analysis was choosing the best pre-processing
technique for GC data. As the GC peaks were discrete variables with large
variation between them regarding fatty acid levels autoscaling was warranted
as the best option. Afterwards, a preliminary data analysis using PCA with the
98 samples and the 7 fatty acid concentration was performed. Samples U14
and U51 were detected as outliers through the residual vs. T2 plot. A deeper
observation of these two samples on the original data reveal a clear reason
for this fact. Both layer have an empty value on their fatty acids (16:1 for U14
and 18:2 for U51), thus, and keeping this fact in mind samples were not
removed.
23
Figure 6. Scores and loading from PCA with 98 samples (Yellow samples belongs to internal subcutaneous fat layer) and 7 fatty acid concentration as variables.
Figure 6 shows clear differentiation between external and internal layer.
As it is known from the literature, the external layer on the subcutaneous fat
has a bigger amount of unsaturated fatty acids. The closer distance with the
environment (colder than the inside of the body) does mandatory the bigger
amount of unsaturated carbons presented on the fat to ensure its fluidity. In
the other hand the fat on the internal layer is always on a higher temperature
(closer to normal body temperature) so the presence of unsaturated fatty
acids is not necessary. As it is clear on Figure 6 PC#2 explain almost the
variability between layers. Its negative values are related to the internal layer
and saturated fatty acids as 18:0 and 16:0 whenever its positive values are
related to the external layer and unsaturated fatty acids. The internal layer is
correlated mainly to 18:0 and the external layer correlated to 16:1.
After this PCA analysis, a PLS was calculated using the iodine values
from the NITFOM as Y on the model. From the three different iodine values,
NITFOM was used for calculation of the PLS model, as NITFOM is less
correlated to fatty acids concentration, than the other two iodine values, which
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
-1
-0.5
0
0.5
1
C18:1
C16:0
C14:0
C16:1
Loadings PC#1 (48.473%)
C18:0
C18:2
C18:3
Load
ings
PC#
2 (2
2.60
6%)
-4 -3 -2 -1 0 1 2 3 4 5
-3
-2
-1
0
1
2
3
4
U51
U61 U87
U75
U99
U55
U66
U81
U85
U83
D83 U91
U74 U86
U95
U68
D61 U92 U73 U93
D81 D86 D91 D75
D51
U90
U59
D87
U89
D59
U42
U38
U58
D85
D99
D55
U43
D73
U36 U31
PCA Scores and Loadings [Model 8]
Scores PC#1 (48.473%)
D74
U49 U80 U47
U35
D68
U33
D93 D58
D66
D89
U40 U29 D90 D92
D95
U17
U12
U22
U19
D80 D43
U16
U39
U26 U32
U44
D49 D47 D40
U25
D42 U14
U28 U34
D38 D25
D26 U10 D31
D44
D28 D14
U37
D29 D19 D33
D12
D36
D10
D35 D17 D16 D39
D22 D34
D32
D37
Scor
es P
C#2
(22.
606%
)
24
are calculated from the fatty acids concentration. Scores plot shows similar
results as the PCA calculated before, with clear differentiation between layers
(Not shown).
Figure 7 shows the NITFOM placed on the loadings close to C18:2 and
C18:3 dots as it is expected from the theory. However, it is important to
observe position of oleic acid 18:1 placed away from the NITFOM dot, which
means that the values are not positively correlated. As it was shown in the
theory, C18:1 and NITFOM should be positive correlated due to iodine value
is a calculation depending on the number of double bonds in fatty acids.
However, the position of C18:1 on the loadings plot could be explained by the
fact that all the fatty acids concentration are strongly correlated between
themselves, as they are explained on percentage. Therefore, what is
explained on the loadings plot is the fact which for the same iodine value the
increase on C18:1 concentration will decrease the concentration of C18:2.
Thus, due to the positive correlation between iodine value with both fatty acids
is much stronger on C18:2 than on C18:1, this situation induces the negative
correlation between C18:1 and iodine value, and it locates oleic acid far from
both, linoleic acid and iodine value. This positioning of C18:1 regarding C18:2
and iodine value is always repeated for the different models calculated and,
for the different iodine values used on the calculation.
Figure 7. PLS loadings using both layers for calculation.
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
C18:2
Iodtal (NitFom)
C18:3
C18:0
C16:1
Loadings PC#1 (46.776%)
PLS Loadings [Model 9]
C14:0
C18:1
C16:0 Load
ings
PC#
2 (2
2.70
8%)
25
Afterwards, different models were calculated in order to find the best
model for iodine value calculation. Thus, 6 models were performed as shown
as follow:
Table 2. Comparative between PLS models calculated using GC-‐MS data.
Table 2 shows the 6 PLS model calculated and their root mean square
error cross-validated and coefficient of determination respectively. As we can
see the model calculated using the internal layer is for both iodine values
better than the model using both layers and only external. Thus, the internal
layer seems to be better on the calculation of iodine value through the FA
composition. The variation between the models using the iodine value GC-
FAME calculated and the models using the iodine value calculated through
the NITFOM is explained by the correlation between both methods of
calculation (R2 = 0,9535). As the iodine value calculated with the NITFOM is a
prediction itself, models using the iodine value calculated with the GC-FAME
are more robust from a theoretical point of view. The fact that better models
are calculated when using the NITFOM could be explained accepting that the
NITFOM prediction removes some variables on its calculation.
4.2. HF-1H-NMR analysis
On a first view of the spectra resulted of the HF-NMR analysis it is
possible to easily identify different functional groups from the fat samples. The
relation between Table 3 and Figure 8 is clearly observable.
26
Table 3. Situation of the peak and functional group related on HF-‐NMR spectroscopy
Figure 8. Raw data obtain from all the HF-‐NMR samples analysis.
The peaks on 4,10; 4,31 and 5,24 ppm are related to the glycerol
molecule from the fat.
Mean centering was the preprocessing chosen for NMR data. A PCA
was carried out. On Figure 9 is observed a quiet clear differentiation between
layers along PC#2 even when some samples are not following this rule.
These samples will be carefully observed on the incoming PLS models. PC#1
and PC#2 explain 88% of the total variance.
0123456-1
0
1
2
3
4
5
6
7
8
9x 106
ppm
27
Figure 9. Score plot for PC1 and PC2 from the NMR results
As explained before, the external layer has a bigger content on
unsaturated fatty acids regarding the internal. Differences between samples
from external layer within samples from the internal layer are visible on the
raw spectra as well.
On Figure 10 it is noticeable the intensity differences between both
spectra. The peak representative for the functional group (CH2)n (around 1.30
ppm) is stronger on the sample from the internal layer sample (D34 green
color), whereas on the peak for CH=CH-CH2 (around 2.04 ppm), the stronger
signal comes from the external layer sample (U34 blue color).
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2
x 107
-1
-0.5
0
0.5
1
1.5
x 107
D42
U87
D40
U12
U74
U43
D59
U39
U51
D86
U29
D87
D35
U38
D10 D49
U93
U25
D36
U90
U92
U22 D26 D39
U91
U33
U28 D44 U83
D34 D81
U40 D29
U95
U35
U86
D47
D89
U37 U36
D51
U16
D38
U44 U26 U34
U89
U66
U31
D55 D25 U47
U73
D85
D33
U14
U42
D80
D68
D75 D22 D92
U55
D17
D83
D93
U85
U81
Scores PC#1 (58.241%)
PCA Scores [Model 2: Mean centered/all variables/all sam ...]
U10
U19
D14
D66 U32
U99
D12 D58
D99
D28 D61 D73
D16 D43 D91
D74
U80
D95 D37
D19
U61
U58
D90 D32
U49
U75
D31
U17
U59 U68
Scor
es P
C#2
(30.
081%
)
28
Figure 10. Detail from raw spectra showing differentiation between subcutaneous fat layers. Internal layer sample in green color and external layer sample in blue color.
In order to evaluate the potential to predict the iodine value, several PLS
models were calculated with both fat layer together and separately. Variable
selection was performed in the different cases using regression coefficient
values where variables with value close to zero were removed (Variable
Importance in Projection method was also used with worse results). Several
tests eliminating variables ranges related to different peaks from the NMR raw
data were performed and no improvements on the prediction were observed.
Both iodine value calculated with the GC-FAME (Calculated from
CAROMETEC fatty acid analysis and calculated with the analysis previously
performed) were selected in different models. The iodine values from
CAROMETEC did not make layer differentiation; hence, more accurate
models were calculated using iodine values calculated from the GC-FAME
previously performed. In the models where the % of fatty acid was included as
variable, iodine values obtained from CAROMETEC were used due to the
direct correlation between the fatty acid % and the iodine value calculated
from the previous GC-FAME. This direct correlation improve falsely the model
being important to avoid that situation.
1.21.31.41.51.61.71.81.922.10
1
2
3
4
5
6
x 106
ppm
29
Outlier detection was performed using the high leverage (T2) values but
no samples were deleted. Results from the different model calculations are
shown on the Table 4 and explained below.
Table 4. Main characteristics of the PLS models calculated from the NMR data.
Model 1 - A first model with all the variables and the 98 samples was
calculated.
Model 2 – Variable selection using Regression Coefficient values was
performed. Variables with value close to zero were deleted from the model
progressively until model started to decrease quality of predictions.
Model 3 & 4 – Calculation with the different layers separately. The
prediction using the internal fat layer seems to be much more accurate than
with the external layer or both layers, as it was supported for the GC data
analysis performed before.
Model 5 – Again variable selection using Regression Coefficient values
was performed. The prediction calculated is quite good, but the high number
of PCs necessary for prediction was encouraging for continuing a better
prediction model search.
Model 6 – The % of fatty acid obtain from the GC-FAME experiment was
added. As explained before, due to the direct correlation between these
values and the iodine value calculated, it was mandatory the use of the iodine
value obtained from CAROMETEC.
The number of PCs used for prediction decrease significantly so new
model using the internal layer data is following. was attained through variable
selection (using coefficient regression again) from Model 5. On each variable
30
selection performed the variables were removed little by little until the model
started to decrease quality by increasing the RMSECV and decreasing R2.
Models 7 – The model was attained through variable selection (using
coefficient regression again). The number of PCs used for prediction
decrease significantly so new model using the internal layer data is following.
Model 8 – Same procedure than in Model 6 was done but using only the
internal layer. The model improves significantly its quality as number of PCs
used decreases, prediction error decreases and coefficient of determination
increases.
Model 9 - It was reached by variable selection using coefficient
regression from the Model 8.
Model 10 – As the iodine value from CAROMETEC was calculated from
the mixing of both layers, a sum of spectra from each layer was performed on
each sample and new model was calculated. This technique was neither a big
difference on the model quality.
Model 11 – Variable selection from Model 10 was performed with slightly
model improving.
Based on these models calculated using the NMR data and the previous
models calculated using the GC-MS, it could be stated that best predictions
models are the ones using only data from the internal subcutaneous fat layer.
The best model using GC-MS has a RMSECV = 2.488 and R2 = 0.8696
(Model 3 in Table 2), the best model gotten from the HF-1H-NMR has a
RMSECV = 1.768 and R2 = 0.9003 Model 5 in Table 4), and the best model
developed using data from both methods has a RMSECV = 2.491 and R2 =
0.8704 (Model 9 in Table 4). These results combined with the fact that NMR
measurements are less time consuming than GC-MS analysis shows HF-
NMR as the best analytical method for iodine value prediction.
31
5. CONCLUSION
The present study shows a clear improvement on iodine value
predictions when using analysis information from the inner subcutaneous fat
layer. This statement is supported by both methods used in the study: Gas
chromatography and HF-1H-NMR.
The different models developed for iodine value prediction, using both
methods separately and together, conclude that the best model is based on
HF-1H-NMR analysis, resulting in RMSECV = 1.768 and R2 = 0.9003 model
parameters.
6. ACKNOWLEDGEMENTS
I would like to thank Bente Pia Danielsen for her incalculable help on the
laboratory work, with really kind and warm assistance in every moment. Also,
I want to thank Josefina Gonzalez-Solveyra for her enormously support on the
most stressful moments, knowing how to touch the piano keys to create the
necessary melodies for appeasing my being on each moment of need. I would
like to thank my family as well, because without their unbroken support I
would not be finishing this step, living as I live, and writing this with the
feelings I am doing it. To all my friends who have been upholder so many
times to my struggle and worries, particularly to Manuel Tobajas and Clara De
Juanas for taking care of my mind even when they have to be focus on the
great experience they are living. I would like to finally thank to the beautiful
nights on this city, for being always the good shelter I need. Gracias.
The hallway ends and all the doors have been left open.
32
6. LITERATURE
AOCS Recommended Practice Cd 1c-85 (2009). Method for calculated iodine value from fatty acid composition.
D.U. Ahn, S. Lutz, J.S. Sim (1996). Effects of dietary alfa-linolenic acid on the
fatty acid composition, storage stability and sensory characteristics of pork loin. Meat Science 43, 291-299
L. Bosch, M. Tor, J. Reixach, J. Estany (2012). Age-related changes in
intramuscular and subcutaneous fat content and fatty acid composition in growing pigs using longitudinal data. Meat Science 91, 358-363
E.A. Bryhni, N.P. Kjos, R. Ofstad, M. Hunt (2002). Polyunsaturated fat and
fish oil in diets for growing-finishing pigs: effects on fatty acid composition and meat, fat, and sausage quality. Meat Science 62, 1-8
S. Christophoridou and P. Dais (2009). Detection and quantification of
phenolic compounds in olive oil by high resolution 1H nuclear magnetic resonance spectroscopy. Analytica Chimica Acta 633, 283-292
F. Dalitz, M. Cudaj, M. Maiwald, G. Guthausen (2012). Process and reaction
monitoring by low-field NMR spectroscopy. Progress in Nuclear Magnetic Resonance Spectroscopy 60, 52-70
A. Daza, C.J. Lopez-Bote, A. Olivares, D. Menoyo, J. Ruiz (2007). Age at the
beginning of the fattening period of Iberian pigs under free-range conditions affects growth, carcass characteristics and the fatty acid profile of lipids. Animal Feed Science and Technology 139, 81-91
A. Daza, J. Ruiz-Carrascal, A. Olivares, D. Menoyo, C.J. Lopez-Bote (2007).
Fatty acids profile of the subcutaneous backfat layers from Iberian pigs raised under free-range conditions. Food Science and Technology International 13, 135-140
L.F. Gladden (1995). Industrial applications of nuclear magnetic resonance.
The Chemical Engineering Journal 56, 149-158 K.G. Grunert (2006). Future trends and consumer lifestyles with regard to
meat consumption. Meat Science 74, 149-160 F.R. Huang, Z.P. Zhan, J. Luo, Z.X. Liu, J. Peng (2008). Duration of dietary
linseed feeding affects the intramuscular fat, muscle mass and fatty acid composition in pig muscle. Livestock Science 118, 132-139
N.B. Kyriakidis and T. Katsiloulis (2000). Calculation of Iodine Value from
Measurements of Fatty Acid Methyl Esters of Some Oils: Comparison with the Relevant American Oil Chemists Society Method. JAOCS 77, 1235-1238
33
S.H. Lee, J.H. Choe, Y.M. Choi, K.C. Jung, M.S. Rhee, K.C. Hong, S.K. Lee,
Y.C. Ryu, B.C. Kim (2012). The influence of pork quality traits and muscle fiber characteristics on the eating quality of pork from various breeds. Meat Science 90, 284-291
L. Mannina, A.P. Sobolev, S. Viel (2012). Liquid state 1H high field NMR in
food analysis. Progress in Nuclear Magnetic Resonance Spectroscopy 66, 1-39
D. Martin, T. Antequera, E. Gonzalez, C. Lopez-Bote, J. Ruiz (2007).
Changes in the fatty acid profile of the subcutaneous fat of swine throughout fattening as affected by dietary conjugated linoleic acid and monounsaturated fatty acids. Journal of Agricultural and Food Chemistry 55, 10820-10826
J. Mitchaothai, C. Yuangklang, S. Wittayakum, K. Vasupen, S.
Wongsutthavas, P. Srenanul, R. Hovenier, H. Everts, A.C. Beynen (2007). Effect of dietary fat type on meat quality and fatty acid composition of various tissues in growing-finishing swine. Meat Science 76, 95-101
Y. Miyake, K. Yokomizo, N. Matsuzaki (1998). Rapid determination of iodine
value by 1H Nuclear Magnetic Resonance Spectroscopy. JAOCS 75, 15-19
K. Nuernberg, K. Fischer, G. Nuernberg, U. Kuechenmeister, D. Klosiwska, G.
Eliminowska-Wenda, I. Fiedler, K. Ender (2005). Effects of dietary olive and linseed oil on lipid composition, meat quality, sensory characteristics and muscle structure in pigs. Meat Science 70, 63-74
H.T. Pedersen, H. Berg, F. Lundby, S.B. Engelsen (2001). The multivariate
advantage in fat determination in meat by bench-top NMR. Innovative Food Science & Emerging Technologies 2, 87-94
T. Pérez-Palacios, J. Ruiz, D. Martín, E. Muriel, T. Antequera (2008).
Comparison of different methods for total lipid quantification in meat and meat products. Food Chemistry 110, 1025-1029
T. Pérez-Palacios, J. Ruiz, I.M.P.L.V.O. Ferreira, C. Petisca, T. Antequera
(2012). Effect of solvent to sample ratio on total lipid extracted and fatty acid composition in meat products within different fat content. Meat Science 91, 369-373
N. Prieto, D.W. Ross, E.A. Navajas, G.R. Nute, R.I. Richardson, J.J. Hyslop,
G. Simm, R. Roehe (2009). On-line application of visible and near infrared reflectance spectroscopy to predict chemical-physical and sensory characteristics of beef quality. Meat Science 83, 96-103
34
J. Ruiz and C. López-Bote (2002). Improvement of dry-cured ham quality by lipid modification through dietary means. In: Research advances in the quality of meat and meat products. Ed. Fidel Toldrá. Research Signpost, Trivandrum, India pp: 255-271 (ISBN: 81-7736-125-2)
J. Segura, C.J. Lopez-Bote (2014). A laboratory efficient method for
intramuscular fat analysis. Food Chemistry 145, 821-825 C. Siciliano, E. Belsito, R. De Marco. M.L. Di Gioia, A. Leggio, A. Liguori
(2013). Quantitative determination of fatty acid chain composition in pork meat products by high resolution 1H NMR spectroscopy. Food Chemistry 136, 546-554
S. Sorapukdee, C. Kongtasorn, S. Benjakul, W. Visessanguan (2013).
Influences of muscle compositionad structure of pork from different breeds on stability and textural properties of cooked meat emulsion. Food Chemistry 138, 1892-1901
H. Todt, G. Guthausen, W. Burk, D. Schmalbein, A. Kamlowski (2006).
Water/moisture and fat analysis by time-domain NMR. Food Chemistry 96, 436-440
N. Viereck, K.M. Sørensen and S.B. Engelsen (2012). Investigating depth
profiles from porcine adipose tissue by HR MAS NMR spectroscopy. Magnetic Resonance in Food Science. ISBN: 978-1-84973-634-3
J.D. Wood, G.R. Nute, R.I. Richardson, F.M. Whittington, O. Southwood, G.
Plastow, R. Mansbridge, N. da Costa, K.C. Chang (2004). Effects of breed, diet and muscle on fat deposition and eating quality in pigs. Meat Science 67, 651-667
top related